ArticlePDF Available

A Survey of Single and Multi-UAV Aerial Manipulation


Abstract and Figures

Aerial manipulation has direct application prospects in environment, construction, forestry, agriculture, search, and rescue. It can be used to pick and place objects and hence can be used for transportation of goods. Aerial manipulation can be used to perform operations in environments inaccessible or unsafe for human workers. This paper is a survey of recent research in aerial manipulation. The aerial manipulation research has diverse aspects, which include the designing of aerial manipulation platforms, manipulators, grippers, the control of aerial platform and manipulators, the interaction of aerial manipulator with the environment, through forces and torque. In particular, the review paper presents the survey of the airborne platforms that can be used for aerial manipulation, including the new aerial platforms with aerial manipulation capability. We also classified, the aerial grippers and aerial manipulators based on their designs and characteristics. The recent contributions, regarding the control of the aerial manipulator platform, is also discussed. The environment interaction of aerial manipulators is also surveyed, which includes, different strategies, used for endeffectors interaction with the environment, application of force, application of torque and visual servoing. A recent and growing interest of researchers about the multi-UAV collaborative aerial manipulation was also noticed, and hence different strategies for collaborative aerial manipulation are also surveyed discussed and critically analyzed. Some key challenges regarding outdoor aerial manipulation and energy constraints in aerial manipulation are also discussed.
Content may be subject to copyright.
A survey of single and multi-UAV aerial manipulation
Abdullah Mohiuddin1, Taha Tarek2, Yahya Zweiri3,1, Dongming Gan1,
Abstract Aerial manipulation has direct application
prospects in environment, construction, forestry, agriculture,
search, and rescue. It can be used to pick and place objects
and hence can be used for transportation of goods. Aerial
manipulation can be used to perform operations in environ-
ments inaccessible or unsafe for human workers. This paper
is a survey of recent research in aerial manipulation. The
aerial manipulation research has diverse aspects, which include
the designing of aerial manipulation platforms, manipulators,
grippers, the control of aerial platform and manipulators,
the interaction of aerial manipulator with the environment,
through forces and torque. In particular, the review paper
presents the survey of the airborne platforms that can be used
for aerial manipulation, including the new aerial platforms
with aerial manipulation capability. We also classified, the
aerial grippers and aerial manipulators based on their designs
and characteristics. The recent contributions, regarding the
control of the aerial manipulator platform, is also discussed.
The environment interaction of aerial manipulators is also
surveyed, which includes, different strategies, used for end-
effectors interaction with the environment, application of force,
application of torque and visual servoing. A recent and growing
interest of researchers about the multi-UAV collaborative aerial
manipulation was also noticed, and hence different strategies for
collaborative aerial manipulation are also surveyed discussed
and critically analyzed. Some key challenges regarding outdoor
aerial manipulation and energy constraints in aerial manipu-
lation are also discussed.
KEYWORDS Aerial platforms, Aerial manipulation, Aerial
manipulators, Aerial grippers, Multi-UAV collaborative trans-
Aerial manipulation, in general, is referred to as an activity
performed by aircrafts with hovering capability, for grasping,
transporting, positioning, measuring by using an end effec-
tor or a gripper, attached to the aircraft [1]. Development
of Unmanned Aerial Vehicles (UAVs) with manipulation
capabilities has attracted high attention in recent years.
UAVs with manipulators are useful for many applications
such as remote inspection [2], [3], cutting high tension
cables [4], packages delivery [5] and monitoring of hostile
environments [6]. Aerial manipulators can perform tasks,
where human access is limited such as; turning a valve [7]
in an inaccessible location, performing the inspection on
bridges [8]. UAVs with manipulators can also be useful
*This publication is based upon work supported by the Khalifa University
of Science and Technology under Award No. RC1-2018-KUCARS.
1KUCARS, Khalifa University of Science and Technol-
ogy, P.O. Box 127788, Abu Dhabi, United Arab Emirates.
2Algorythma’s Autonomous Aerial Lab, Abu Dhabi, UAE.
3Faculty of Science, Engineering and Computing, Kingston University
London, London SW15 3DW, UK.
for canopy sampling [9]. Recently a UAV-manipulator was
proposed to place a wrist band in the hands of a victim for
search and rescue [10]. Researchers are also considering the
collaboration of aerial manipulators with ground robots [11].
Aerial platforms are being considered for their potential
ability to transport goods, quickly and cost-effectively. A
recent study showed that drone delivery could help reduce the
greenhouse gases [12]. Aerial platforms can also transport
deformable linear objects such as hose, or ropes [13]. A
demonstration of building a rope bridge using multi-rotors
is shown in [14].
Aerial manipulation research benefits from the existing
knowledge of manipulator designs and development of aerial
platforms such as multi-rotors and helicopters. However, the
design of manipulators and controllers for aerial platforms
requires some special considerations which are typically
not needed in ground-based manipulation and transportation.
Specifically, these considerations include the stability of the
aerial platform, since any motion of the manipulator or any
interaction of the manipulator with the environment will
directly affect the flight of aerial platform. Aerial platforms
are also constrained to use low weight manipulators since any
added weight increases the thrust requirements and the power
consumption. All the above constraints require researchers to
focus on multiple aspects of aerial manipulation.
A. Motivation and scope
The motivation for this paper is to benchmark the state of
the art in aerial manipulation research. A review of multiple
aspects of aerial manipulation would not only help profes-
sionals inside the field but also the professionals outside
of the field to familiarize with state of the art in aerial
manipulation and may also help them apply some of these
methods to their specific application areas. The scope of
this survey is based on the current active research areas in
aerial manipulation which include, research in development
of new aerial platforms for aerial manipulation, or improving
the existing platforms, designing of aerial manipulators and
grippers, improving the control of aerial platform and the
manipulators, the interaction of the aerial manipulator with
the environment and the collaborative aerial manipulation
using multiple UAVs.
B. Research groups involved in aerial manipulation
The growing interest of researchers in aerial manipula-
tion is also, evident from the increasing number of re-
search groups involved in aerial manipulation. Some research
groups and their areas of research in aerial manipulation and
their active aerial manipulation projects are summarized in
Autonomous Systems, Control and Optimization Lab
Johns Hopkins University
Quadrotors with manipulators Aerial grasping
Autonomous Systems Lab
ETH Zurich
Quadrotors Helicopters Aeroworks, AiRobots, Reely, myCopter
Aerial grasping and manipulation
Yale University
Helicoptor with gripper,
quadrotors Design and control
Aerial Robotics
KU Leuven
Innovative rotary platform
Aerial platform development
Design and control, paylaod transportation
CargoCopter project
Drexel Autonomous Systems Lab
Drexel University bjr37/
Quadrotors 2-4 DOF dual-arm Control
German Aerospace Center Helicopter 7 DOF arm Aerial manipulation
Carnegie Mellon University
Quadrotors Geometric control of dynamic
Aerial manipulation, multi-UAV aerial manipulation
Institute of Robotics and Intelligent Systems
ETH Zurich
Helicoptor with gripper,
Multi-UAV co-operative aerial
manipulator, multirotor design
Institute of Robotics and Mechatronics Helicopter with arm manipulator Aerial manipulation for
Interactive & Networked Robotics Laboratory
Seoul University
SmQ platform,ODAR Aerial platform development
Jouhou System Kougaku Laboratory (JSK)
University of Tokyo
Transformable multi-rotor Aerial platfrom development
MultiScale Robotics and Automation Lab
Purdue University
Aerial platfrom development
University of Naples
Quadrotors with serial link
Development of aerial manipulation systems
for industry with ongoing project AeroArms
Robotics Vision and Control Group
Universities of Seville
Pablo de Olavide University
Aerial robotic manipulation
involving control, perception and planning
Current projects: Aerial Robot Coworker
Robotics and Mechantronics group
University of Twente
Aerial manipulator Control, ongoing project: Aeroworks,
SIRSLab - Siena Robotics and Systems Lab
University of Siena
Quadrotors Aerial grasping
Robotics, Intelligent Systems & Control (RISC)
Hexrotors, quadrotors Passive aerial grasping
Vijay Kumar Lab (GRASP)
University of Pennsylvania
Micro-aerial vehicles Multi-UAV co-operative aerial
(a) (b) (c)
Fig. 1. Examples of a) multi-rotor with manipulator [15] b) helicopter with manipulator [16] c) simulation platform for aerial manipulation [17]
Table I. This list does not contain all the research groups
currently active in the aerial manipulation but provides an
idea to the reader about the people active in this field.
C. Structure of the survey
The paper is structured as follows: section II covers a
brief discussion of aerial platforms, section III is about
grippers and manipulators, section IV is presenting planning
and control, section V is a survey of multi-UAV aerial
manipulation, while section VI is regarding challenges of
outdoor aerial manipulation and section VII is a review
of the energy constraints of aerial manipulation. Some key
observations and open research problems are discussed in
the section VIII which is followed by the conclusion in
section IX.
Several aerial platforms with autonomous flight capability
are currently available, such as multi-rotors, helicopters,
fixed winged planes, but not all of them are well suited for
aerial manipulation. Helicopters and multi-rotors have hov-
ering ability which is not available with fixed winged planes.
Some novel aerial platforms are also recently developed with
hovering capability and promising prospects in aerial manip-
ulation. Hovering capability allows an aerial platform to have
a relatively fixed position in air and manipulator attached to
the aerial platform can perform aerial manipulation tasks. A
brief discussion of these aerial platforms which are used in
aerial manipulation is provided in following subsections II-
A II-B II-C II-D. The next subsection II-A will discuss multi-
rotors, which are the most common aerial platform used
in aerial manipulation research. Subsection II-B is about
helicopters, and after that subsection II-C will introduce
some novel aerial platforms with good potential for aerial
manipulation. The last subsection II-D will present some
simulation platforms that are used for aerial manipulation
A. Multi-rotors
Multi-rotors are most commonly discussed aerial plat-
forms for aerial manipulation [18], [19], [20], [21]. The
hovering capability present in multi-rotors provides an oppor-
tunity for the UAV to remain airborne while any manipulator
attached to it, can perform a useful operation. Multi-rotors
come in various configurations depending on the number
of rotor arms, the orientation of rotor arms, number of
propellers per arm, propeller configurations. Regarding the
number of arms, the most popular are quad-rotors, hex-rotors,
and oct-rotors. Based on rotor arm orientation, the most
popular configuration is the cross configuration as compared
to a plus configuration. The propeller configurations [22]
can be either single propeller per arm or co-axial config-
uration, which means two propellers per arm. In single
propeller per arm configuration, propellers can be placed
above or below the arms. Multi-rotors are popular amongst
the hobbyists and are commercially available; however, they
have some constraints, when considering multi-rotors for
aerial manipulation. Commonly used multi-rotors are under-
actuated systems, where translation motion depends on atti-
tude control and any attempt in attaching a manipulator to
aerial manipulator, must take the effect of manipulator on
attitude control into account. In multi-rotors, the rotational
direction of all rotors is kept such as to achieve a net zero
reaction torque, which is obtained by rotating pairs of rotors
in opposite directions canceling out each other’s torque. The
yaw motion is achieved by slowing down the opposite pairs
of motors relative to the other pair. The yaw motion can
be used to apply torque by attaching a gripper on top or
below the multi-rotor. However, the amount of torque that
can be applied is limited by the rotor capabilities and the
length of the rotor arm. Another constraint with multi-rotors
is the difficulty in scaling up. In simple terms, scaling up
a multi-rotor with currently available technology is difficult,
since the thrust produced by the rotors is increased by a
square of the sweeping area of the rotors. Increasing the
sweeping area of rotors requires the increase in the size
of the quad-rotor which means the volume will increase.
The weight or volume increases by the cube law. This
simple analysis means that at some point scaling will not
work. In general, rotor-crafts can be scaled up to a certain
size [23]. It is also argued by [24] that scaling up a multi-
rotor could also be dangerous and more expensive. Although,
recent announcements of a large multi-rotor called Ehang is
developed, till date no commercial or passenger flights are
operated, or any data regarding the flight stability is provided.
It is also important to note smaller quad-rotors are more
agile [25]. The loss of agility while scaling up a multi-rotor
might also result in less control or less ability to compensate
for the inertia change due to the manipulator motion. Another
constraint with multi-rotors as an aerial manipulator, is the
reduced flight time, with a typical flight time of 15 to 30
minutes [26]. The payload carrying capacity is also limited.
Addition of a payload, also adversely affects the flight time.
In general unilateral Electronic Speed Controllers (ESCs)
are used for the motors of the multi-rotors. Multi-rotors
with unilateral ESCs cannot exert downward force larger
than their weight [27]. A manipulator can be attached to
a multi-rotor as shown in Figure 1(a). Further discussion
of design and control of multi-rotors with manipulators is
discussed in sections III-B and IV-B. Recently [28] presented
the design of fully actuated multi-rotor called as Tilt-Hex.
Tilt-Hex provides better pose control. It can also apply force
and torque independently with the help of a rigidly fixed
manipulator [28]. As of today, no studies were found in the
literature where a moving manipulator is attached to a Tilt-
Hex. This review will mainly focus on under-actuated multi-
rotors and helicopters with manipulators.
B. Helicopters
Helicopters are also widely referred, in aerial manipulation
literature after multi-rotors [29], [30], [31], [32], [16], [33],
[34]. They also have hovering capability, like multi-rotors,
but better payload capacity as compared to multi-rotors. He-
licopters can be easily scaled up in size and payload capacity
as compared to multi-rotors. Examples of helicopters with
better payload capacity in the literature which enables them
to fly with relatively heavy arm manipulator are [35], [16].
An example of a helicopter with an arm manipulator is
shown in Figure 1(b). Unlike multi-rotors, a conventional
helicopter consists of a single main rotor and a tail rotor. The
main rotor is responsible for creating enough thrust for the
helicopter to lift it. The translational motion of a helicopter
is achieved by creating differential lift, by cyclic control
of the angle of attack of the rotor blades. The ascending
and descending motion can also be achieved by a collective
change of angle of attack of both rotors, thereby changing the
thrust produced. The rotation of single main rotor causes, a
reaction torque, which is countered by the tail rotor. The
yaw motion of a helicopter is achieved by increasing or
decreasing the thrust produced by the tail rotor. The thrust
produced by the tail rotor is adjusted by varying the angle
of attack of the blades. The yaw motion generated by the
tail rotor could result in a higher torque since the length of
the tail adds to the yaw torque. This could be useful when
a gripper is attached just below the helicopter and needs
to perform an operation requiring high torque. Furthermore,
the efficiency of a single but bigger rotor is higher than
multiple small rotors. One big rotor can have more chances
of creating a blowing effect, i.e., disturbing the object that
is to be picked. One big rotor also means more chances
of collision of blades while performing aerial manipulation
so maybe a bigger manipulator is needed for any lateral
manipulation. The main rotor also causes oscillation effects
which makes a helicopter challenging to operate. Examples
of these oscillation effects are mechanical resonance in air
and ground resonance [36]. Furthermore, current helicopter
models used in literature for aerial manipulation do not
have any protective ring around the main rotor, which is
sometimes present in multi-rotors and makes it safer for
aerial manipulation. Blade flapping is expected to be higher
in helicopters since the main rotor is longer hence more
bending near the ends of the rotor. Aerodynamics of multi-
rotor and helicopters are also different; it is obvious that a
helicopter would be susceptible to more lateral drag because
of the more lateral area. Hence wind disturbances will
cause more positioning errors. It is not possible to attach a
manipulator on top of helicopters, whereas it is possible and
practiced in multi-rotor [37], [7], [38]. This also prohibits
the use of helicopters for contact inspection under the roof
like surfaces such as bridges. Helicopters are also generally
less agile and have less maneuverability in tight spaces [39];
thus the addition of an arm manipulator would also impact
its agility and maneuverability. Addition of arm manipulator
also makes the system more sensitive to the vibration and
resonance effects caused by the main rotor [36].
C. Novel aerial platforms
Most of the studies presented in the literature are using
commercial UAVs for the aerial manipulation research. How-
ever, researchers have also proposed, novel platforms [47],
[27], [44], modification to existing platforms [40], [48], [49],
whereas [46] presented a technique of designing a new aerial
platform based on application. Various new trends are being
observed, one of them is exploring the possibility of hybrid
aerial vehicles which can move on various mediums such
as water and ground along with flight capabilities [42], [3],
[49] and the other trend is creating re-configurable vertical
take-off and landing aircrafts (VTOLs), which do not have
fixed shapes [45], [44]. An account of all these is discussed
The modification was proposed by [40] in the form of the
addition of a buoyant structure to the quad-rotor to overcome
the instability of the flight for redundant manipulators as
shown in Figure 2(a). Researchers planned to attached a
manipulator to the proposed hybrid UAV. The test runs
in [40] show that this hybrid structure without the manip-
ulator does stabilize the flight. However having attached
a buoyant structure such as the one shown in Figure 2(a)
must add the aerodynamic effects such as parasitic drag.
It also increases the lateral surface area which means that
perturbations caused by wind might offset the benefit that is
being achieved by the addition of buoyant structure.
Another novel aerial platform recently presented, is called,
spherically-connected-multi-quadrotor platform (SmQ Plat-
form) [47] as shown in Figure 2(b). Three quadrotors were
attached to small beams using spherical joints. The small
beams were concentrically connected to a small circular
plate acting as the base frame. The connection of multiple
quadrotors to a single frame allows increasing the total
thrust force that can be generated by the aerial platform.
SmQ platform is also able to execute yaw motion and can
apply horizontal force using a link that is extending out
from the base frame. The system although is over-actuated,
but the actuation is limited by the range limit of spherical
joints. Another novel aerial platform recently presented, is
(a) Quadrotor with
bouyant structure [40]
(b) Spherically connected multiQuadrotor [41] (c) Omni-Directional Aerial Robot [27]
(d) Hybrid ground aerial
platform [42]
(e) Hybrid aerial platform for inspection
of wires [3]
(f) A bio-inspired Spider
Micro-Aerial Vehicle [43]
(g) Transformable aerial platform [44] (h) A modular VTOL Vehicle [45] (i) Platform developed us-
ing the method presented in
ref [46]
Fig. 2. Examples of novel aerial platforms discussed in literature
called, omni-directional-aerial-Robot (ODAR) as presented
by in [27] as shown in Figure 2(c). ODAR is a fully
actuated aerial platform with reversible ESCs which can
produce a bidirectional current; hence the ODAR platform
can apply force in both directions; a feature that is currently
not available in conventional commercial multi-rotors and
Researchers [42], [3], [49] have also tried to create a
hybrid platform, which can not only fly but travel through
another medium. One example [42] is shown in Figure 2(d)
in which the aircraft can not only fly and hover in the
air but also travel on the ground, after going through a
transformation, where it folds its arms and forms a two-wheel
vehicle. This kind of hybridization, can increase the reach
of aerial platform further and also save energy by traveling
through the ground when flying is not necessary. Another
example is [49] where the aerial platform can not only
hover and travel underwater, but it is also capable of flying
in the air, during the same flight. In above two cases, the
prospects of attaching a manipulator are yet to be explored,
and integration of a manipulator to such platforms would
require considerations, regarding the travel through another
medium based on the requirements of the other medium.
Another example [3] of an aerial platform traveling through
another medium is shown in Figure 2(e), where the drone can
hover in the air, but it can also land on wires for inspection,
and after landing it can travel on wire, and fly again, in
case of any obstacle on the wire. Another example [43]
is the Figure 2(f), which has an additional mechanism to
launch anchors. These anchors attach to the surroundings
and hence provide support to the UAV. This anchoring or
perching allows the UAV to save energy instead of the usual
hovering in position performed during an aerial manipulation
A novel concept is the use of a transformable aerial
platform, where the whole platform can act as a gripper. In
this case, the aerial platform consists of rotor arms, which
can change their configuration, as compared to fixed rotor
arm multi-rotors [44] as shown in Figure 2(g). In this type
of arrangement, any external manipulator arm or gripper is
not required. In a demonstration, it successfully transformed
itself such that it encompassed the box, grasped it and lifted
it. The transformation of the aircraft also makes it more suc-
cessful in navigation through narrow passages as compared
to other types of multi-rotors, since it can transform its shape
based on the passage. Since the whole platform transforms
and interacts with the object, it seems necessary to add a
protective ring around the rotors to protect the rotors and the
object. This is not necessary for conventional multi-rotor-
manipulator systems since a manipulator adds reachability
and provides some disconnection from the instabilities of
interaction with the object.
A similar concept is having a modular vertical and take
off vehicle [45], as shown in Figure 2(h) where, multiple
individual modules with single rotors, can join together via
magnets to form various configurations. The individual mod-
ules use wheels to move around on the ground and connect
to the other modules, and after forming a configuration, the
modular VTOL can fly.
Another approach could be designing a platform from
scratch, based on the application. Designing a new platform,
instead of using existing ones, could provide more flexibility
in terms of application-specific attributes that might not
be present in existing quadrotors, hex-rotors or helicopters.
These attributes could be a desired force or torque acting on
the end-effector based on any application. Hence a technical
analysis can be performed [46] to determine the mechanical
structure of the system, the number of thrusts and geometry.
A platform that was developed by following the method
described in [46] is shown in Figure 2(i). It can be seen
in the Figure 2(i), that the platform is quite unconventional
in terms of its structure and the location of actuators on the
D. Simulated UAV platforms
A platform that can mimic the aerial behavior of a UAV
can provide a useful tool for the testing of heavy manipula-
tors. Studies performed by [17], [40], [50], [51], [52], instead
of using UAV, used a platform that can mimic the effects of
flight. An example of such a platform is shown in Figure 1(c).
The authors suggest that UAV simulating platform bridges
the research gap between the current aerial platforms lacking
high payload capabilities and thus the capacity for more
dexterous heavy manipulators. A UAV simulator was devel-
oped by [40], [50], which uses a mathematical model of a
quad-rotor to emulate the behavior. A six-degree-of-freedom
miniature version of the gantry crane was used to replicate
the UAV by [17], which provides the complete range of
motion of a rotor-craft as well as ground truth information
without the risk associated with free flight. These UAV test
rigs, however, do not account for various aerodynamic effects
experienced during highly dynamic flying maneuvers.
Although the working principle of manipulators and grip-
pers attached to a ground robot or an aerial robot are the
same, however, the aerial manipulators and grippers require
some considerations such as; aerial manipulators are de-
signed while considering the effects of moving manipulator
on the aerial platform. The grippers for aerial platforms are
designed while considering the effects of positioning errors
of the UAV during grasping of an object. The challenges
imposed by aerial manipulation can be dealt with, via both
improving the design and control of the systems; therefore
researchers have attempted to improve not only the design
of aerial manipulators and grippers but also the control
strategies. This section is divided into two subsections, where
section III-A surveys different types of aerial grippers and
their characteristics and section III-B presents different aerial
manipulators, some design examples from literature and
characteristics of different types of aerial manipulators.
A. Grippers as end-effectors in UAV-manipulator systems
Aerial platforms can carry a wide range of low weight end-
effectors such as NDT (Non-Destructive-Testing) sensors and
grippers. NDT sensors require the design of manipulators
and interaction control strategies to incorporate the effects of
resulting instabilities during contact based inspection. Any
design aspect of NDT sensors themselves does not seem
relevant to aerial manipulation research hence the current
section will focus on grippers attached to the aerial platforms.
The most simple design of gripper for aerial manipulation is
a hook [53], however it will require precise position tracking
of manipulator and the aerial platform. Grippers are the most
common type of end-effector. They usually have different
gripping techniques and actuation styles. Grippers for aerial
manipulators generally include the following characteristics:
light-weight, payload shape compatibility, the reliability of
grasping, the mechanism to deal with reaction forces, ability
to compensate for positioning error of UAV. A comparison
of different characteristics of common grippers used in aerial
manipulation is provided in Table II.
1) Compliant grippers: A mechanically compliant system
deals with position uncertainty of the UAV by incorporating
the compliance in the gripper [54]. This ensures that the
errors in position control of the UAV do not result in large
forces, and the gripper conforms to the object. Grippers
with compliance can be divided into two categories, ac-
tive compliance, and passive mechanical compliance. Active
compliance is based on active control, utilizing sensors and
actuators which work together to get the desired, force-
deflection relation. Passive mechanical compliance, however,
is achieved via attaching springs in robot joints, to create the
allowance for large joint deflections. This results in lower
contact forces.
An example of passive mechanical compliance is a sin-
gle degree of freedom anthropomorphic finger module that
was presented in [55] as shown in Figure 3(a) for aerial
manipulation and grasping. The finger module was driven
by a high torque motor that moves the three joints of the
finger using a tendon. Elastic elements were used to keep
the joints extended or open. This elastic element results
in the passive mechanical compliance of the finger module
against collisions with any objects. The authors extended
their work in [56] by attaching the compliant finger to the
compliant arm, which is further discussed in manipulator
Gripper types Compliance Shape
restriction Manipulate Grasp Destructive
Impactive gripper Low Low None Yes Holds No
Ingressive gripper Low Low Yes (soft surface) No Penetration Yes
Mechanical compliant gripper High High None Low Holds No
Magnetic gripper None Planar/curvature surface Yes (Ferrous) Low Magnetism No
Vacuum gripper Low Medium None Low Self sealing suction No
design section III-B.1. Another means of compliance in grip-
pers can be achieved by adaptive under-actuation. A gripper
with fewer actuators as compared to the degree of freedom
shows adaptive behavior. One approach was from [54] as
shown in Figure 3(c) which employs passive mechanical
compliance via adaptive under actuation in a gripper to allow
for large positional displacements between the aircraft and
the target object. Another example of the under-actuated
gripper is the study by [57] which discussed aspects of robot
hand performance specific to grasping and perching from
an aerial vehicle [57] also explored the impact of design
and grasp parameters including tendon routing/pulley ratio,
object size, and palm-size on the performance of both fully
and under-actuated designs. It was shown that apart from
perching applications, under-actuated designs consisting of
single actuator per finger perform sufficiently, whereas for
perching tasks fully actuated designs perform better.
Shape conformity is also achieved in a proposed aerial
gripper [58], via using a prismatic joint to change the spacing
of the fingers to match with the size of the object. After
matching the size of the object the re-volute joints in the
finger can rotate the finger links to wrap around the object.
An example of active mechanical compliance is the gripper
which is a hybrid of an arm manipulator and gripper [59].
The concept is similar to dual arm manipulator with 2DoF
fingers. The independent control of the fingers allows for
more shape conformity.
2) Ingressive grippers: Ingressive grippers [60], [61] can
grasp the objects without any well-defined attachment points.
As shown in Figure 3(b) these grippers penetrate the metal
hooks into the payload to attach themselves to the surfaces.
That is why this gripper is useful for objects such as wood
which can permit the penetration of hooks. Apart from that,
the object must have enough planar surface area for the hooks
of the gripper to penetrate. Ingressive grippers can cause
some damage to the surface of the payload, which limits
these types of grippers in applications which permit surface
penetration of hooks into the payload. The advantage of this
type of gripper is that it does not require to fully enclose the
object to grasp it, whereas, the impactive grippers require
the gripper to enclose the object for grasping as discussed
3) Impactive grippers: Impactive grippers [60] are those
which use clamping motions to enclose the payload. Im-
pactive grippers also grasp the object by applying sufficient
normal forces or use frictional forces to hold the object;
therefore it is necessary that the geometry of the payload be
compatible with the gripper. It also requires creating enough
friction between the contact surfaces of the object and the
gripper. An example of the impactive gripper is [7], where
an impactive gripper is attached on top of a hex-rotor as
shown in Figure 3(d). The authors exploited the yaw motion
of the hex-rotor platform, to create a torsional force that can
accomplish tasks such as plucking the harvest, taking the
light bulb off. Another example is in [62] where an impactive
gripper grasped a foam brick.
4) Magnetic grippers: The gripping or grasping process
can be simplified by using magnetic grippers for metallic
payloads or payloads with metallic attachment points [63],
[64], [24]. These magnetic grippers are commercially avail-
able and come in various specifications such as the one called
OpenGrab developed by Nicadrone [65]. Some of these grip-
pers use Electro-permanent-magnets which do not require a
constant power consumption to stay activated. The magnets
available in the market, however, do not take into account for
any curvature of the payload surface, therefore in one study,
the authors [63] designed a mechanically compliant structure
that housed magnetic grippers, which enabled it to attach
to surfaces with curvature successfully. Using a magnetic
gripper and magnetic attachment points on payload simplifies
some constraints associated with aerial gripping, on the other
hand, it is also limited due to a magnet‘s load carrying
capability.The biggest drawback of magnetic grippers, is
their limitation to attach to ferromagnetic materials, which
greatly restricts their application range. Magnetic grippers
were also used in a recent Mohamed Bin Zaid International
Robotics Challenge (MBZIRC 2017), where most of the
teams used magnetic grippers to lift a metallic payload. The
gripper developed by KAUST team, [66], [67] consisted of
permanent magnets, and the object was detached by creating
a gap between the permanent magnets and the object using
a plate that is moved by a servo actuator. Magnetic grippers
also accompany sensors such as infrared or camera to detect
the object and to provide confirmation of attachment of
5) Vacuum grippers: A vacuum gripper works by using
suction created in the cup in contact with the payload
surface [68]. Usually this is performed using a vacuum pump.
The vacuum pump used in [68] comprised of 24% of the total
mass of the quadrotor. The vacuum gripper in [68] showed
the ability to successfully grasp a wide variety of objects
including a USB plug, battery, wood block, eye glasses, hair
(a) (b) (c) (d)
Fig. 3. Examples of grippers a) Compliant anthromorphic finger module [55] b) Ingressive gripper, gripping a wooden part [60], [61] c) Adaptive
under-actuated compliant gripper [54] d) An impactive gripper for torque application [7]
brush, rubber duck, box, wood block, metal tin and plastic
container with angled lid. The compliant behavior is added
by using a spring-based universal joint supporting the plate
with suction cups. However, the ability to conform to the
shape of the payload is not available.
6) Electroadhesion based grippers: The principle of elec-
troadhesion [69] is the electrostatic attraction. This type of
attachment works with dirty or dusty surfaces, however the
surface must be smooth. It requires, however, a continuous
supply of small amount of energy to keep the attachment.
Electroadhesion was used successfully as a perching mech-
anism for a micro-UAV by [70]. This method is promising
for smaller UAVs. It is difficult to manufacture the grippers
with actuators for small insect sized UAVs. The method
is electrically and mechanically simple which is helpful
in case of insect sized UAVs. The only requirement is
the electro-adhesive patch and the voltage source. Another
drawback of this method is the dependence of the method on
environmental factors such as humidity, material properties,
current leakage.
B. Aerial manipulators
A manipulator is defined as an object added to the multi-
rotor to extend its reach or reduce disturbances to the
airborne platform during interaction with the environment.
Attachment of a manipulator to a multi-rotor changes the
dynamic behavior of the multi-rotor due to, change of
center of gravity (CoG), the variation of inertia, and effects
of dynamic reaction forces and torques generated by the
movement of the arm [71]. One way to tackle this change
of dynamic behavior is by designing appropriate control
algorithms that can reduce flight instability, and the other
way is to design manipulators that reduce these three causes
of instability. Control methods to deal with these causes of
instability are discussed in the control section IV-B, while
this section is focused on the design of manipulators.
The design of manipulators attached to UAVs is strongly
constrained by the weight of the arm, in fact, one of the
most important design criteria is to improve payload to
manipulator weight ratio. The weight of the manipulators
can be reduced by the 1) selection of low weight material for
manipulators 2) selecting low weight actuators. The designs
of the aerial manipulators are influenced by the requirements
of certain desired characteristics. These desired character-
istics of the manipulators may include, ability to absorb
the impacts, reach-ability, dexterity [72], and compliance.
Certain types of manipulators exhibit different characteris-
tics, which is why it is essential to discuss the types of
aerial manipulators. Aerial manipulators can be classified
into four basic types based on literature survey and the
definition of manipulators as described at the beginning of
this subsection III-B. Link-based serial manipulators, parallel
linked mechanisms, hydraulic manipulators and cable manip-
ulators. Each one of these types has different characteristics
as summarized in Table III, and discussed below.
1) Link-based serial manipulators: Link-based serial ma-
nipulators are widely used in aerial manipulation as shown in
Table IV which is a summary of contributions regarding se-
rial link manipulators. Link-based serial aerial manipulators
generally consist of rigid links, connected via joints which
are actuated by servo motors. The complexity of serial link
manipulators increases with the number of links. Usually,
with every joint, a servo motor is required for actuation.
Thus adding more links can increase inertia disturbances
and cause offset in CoG. An approach in minimizing the
disturbance caused by serial manipulators is by keeping the
motor actuators close to the base of the arm [?], [76]. In one
example of a serial manipulator, the first two DC motors are
included into the base of the arm, while the last two are
positioned at the beginning of the second joint. As shown
in Figure 4(a) belts were used for the motion transmission
for the third joint [?]. Keeping the motors close to the
base constrains the center of gravity of the arm as close
as possible to vehicle base, thus reducing the total inertia
and static unbalancing of the system. The manipulator was
able to fold on itself during landing, or when not in use.
The CoG stays close to the vehicle base in folded position,
which minimizes disturbances, especially during takeoff and
landing. A deployable or foldable manipulator also helps in
keeping the CoG disturbance to a minimum. In [76] the serial
manipulator was based on tentacles based design for which
actuators were close to the base, the motion was transmitted
via cables. A deployable, low mass manipulator for UAV
helicopter was developed by [29]. As shown in Figure 4(b),
it was developed based on steel tubular booms which can
be wrapped around a pulley and deployed when needed.
Apart from the above approach, CoG change caused by
the manipulator motion can be compensated by adjusting
the battery position in the UAV. The battery is a heavy
component of an aerial vehicle, and its position in the UAV
Manipulators Dexterity Reachability Reactionary
of CoG Weight Compliance
on takeoff
Parallel linked
mechanisms High High Low Low Low Number
of links Elastic joints Low
Linked based
serial Higher Higher Higher Higher Higher Number
of links Elastic joints Higher
Cable based None None None None Low Low Good Low
manipulator Low Low Low Low Low High None Low
(a) Arm with motors
close to base, and
motion transmission
with belts [15]
(b) A deployable manipulator [29] (c) Arm with interlocking mechanism for im-
pact absorption [73]
(d) Arm with linear
servo and spring for
compliance [74]
(e) A dexterous manipulator on a
UAV simulator [50]
(f) A small delta manipula-
tor [75]
(g) A hydraulic manipulator [4]
Fig. 4. Examples of manipulators, including serial, parallel link mechanisms and a hydraulic manipulator
can affect the value of CoG of the aerial vehicle. A system
was proposed by [21] which helped in balancing of the CoG
by the short movements of the battery.
In general both serial and parallel linked manipulators are
susceptible to unwanted force or impact that could result in
the transfer of this force to the base of the UAV and hence
cause instability. However, in this section, some examples of
serial link manipulators are presented where researchers have
attempted to reduce these effects. Any unwanted interaction
is usually dealt with by incorporating passive mechanical
compliance. Springs or elastic bands are used to achieve
compliant behavior. As shown in Figure 4(c), [73] proposed
the design of an aerial manipulator, which includes a passive
joint capable of storing impact energy in the form of elastic
potential energy. The impact compresses the elastic band,
while decompression is prevented by using mechanical in-
terlocking. The resulting manipulator can absorb the impacts
without making the UAV-manipulator system bouncing away
from the impact object. Elastic joints were also used by [56],
but without the interlocking mechanism, however, in [56],
elongation of spring in the joints is used to measure torque
and deflection by using a potentiometer. Another example
of elastic joints is [56] in which the joint was driven by a
linear servo as shown in Figure 4(d). The linear servo and
the link were connected via springs. These springs not only
provided compliance but the elongation of springs also helps
in determining the mass of the payload. Above methods of
adding compliant behavior in manipulators are useful, but
the compliant behavior, cannot be altered since it depends on
the elasticity of the springs or elastic elements used. There
is another method of ensuring compliance which is also re-
configurable. This method is based on impedance control
of the end-effector, it is discussed further in the coming
section IV-C.1, under the interaction control.
Serial link manipulators can be designed to have good
reachability and dexterity. A hyper-redundant manipulator
provides good reachability and dexterity and gives the flexi-
bility to control the hovering position of the UAV [72]. It is
also demonstrated that serial link manipulators can become
hyper-redundant by adding more links. It enables the ma-
nipulator to use an extra degree of freedom to grasp objects
which are usually not graspable with its end-effector. The
extra degrees of freedom also help plan the manipulator tasks
in a way that impacts on platform stability are minimized.On
the other hand, having a higher degree of freedom and more
links adds to the overall weight of the manipulator. Some
aerial manipulators with good dexterity, including [72], [17],
[50] were tested while they were attached to test-rigs or
gantry. The paper [17] described the design of a system to
emulate a flying, dexterous mobile manipulator. Two four
degree-of-freedom manipulators were attached to the gantry
system for performing the grasping tasks. Computer vision
techniques and force feedback servoing provide target object
and manipulator position feedback to the control hardware.
Another study by [50] presented a design of a 7 degree
of freedom arm with 4 degrees of freedom of the hand as
shown in Figure 4(e), although the arm was attached to a
test rig capable of mimicking an aerial platform, the authors
claim that this manipulator can successfully manipulate and
transport various objects while maintaining stable flight.
A novel type of serial link manipulators is a protocentric
manipulator. A Protocentric manipulator as defined by [77]
has multiple arms with first joints coinciding with the center
of mass (COM) of the aerial platform. A protocentric ma-
nipulator can be used to grasp cylindrical objects with larger
diameters by enclosing the object using at least two arms. It
also seems possible to keep the inertial parameters similar
by balancing the motion of one arm to the motion of another
2) Parallel linked mechanisms: One way to minimize the
disturbances caused by the aerial manipulators is to use delta
manipulators [75], [87], [88], [19], [89] instead of a serial
manipulator. Delta configuration of manipulators is usually
very lightweight, and robust compared to a serial manipulator
since the actuators are attached to the base [87], [53], [10].
If the delta manipulator is attached to the UAV such that
the force acting on the delta manipulator is passing through
the CoG of the manipulator, it can counter some moments
acting on the aerial platform due to the reactionary forces of
the manipulator. Another example of delta manipulators is
demonstrated in [75] as shown in Figure 4(f), where authors
developed a non-destructive testing end-effector on a three
degree of freedom delta robotic manipulator. The sensor is
mounted on Cardan gimbal joint that allows the sensor to
interact compliantly with the remote environment. Another
delta parallel links manipulator was designed by [88]. The
manipulator was attached at the base of the UAV and is
capable of moving its end effector in the regions below the
UAV and on its sides. The authors used a parameter ”global
conditioning index” to come up with the design values for
the length of the links. Global conditioning index gives a
quantitative measure of the dexterity of the arm. The pro-
posed manipulator is actuated by motors in pitch and planar
directions. This is the only example of a delta manipulator
with good reachability, while other delta manipulators are
usually smaller and are attached on the side of the UAV along
with a sensor and UAV has to move to perform the task.
A delta manipulator causes fewer perturbations to the host
platform, which makes it better than a serial manipulator, but
usually, delta manipulators are smaller with less reachability;
hence the improvement of reachability for delta manipulator
as proposed by [88] proves that delta manipulators with good
reachability can be designed.
3) Hydraulic manipulators: A low weight manipulator is
favorable for aerial platforms, but in the case of a high torque
requirement from an aerial manipulator, the criteria would
be weight to power ratio. Applications such as cutting high
tension cables require a very powerful end-effector which
could provide very high power in a very short time. If electric
motor actuators will be used higher torque can be achieved
by the use of gears but that will also add additional weight
and disturbance from the gearbox. A hydraulic manipulator
as shown in Figure 4(g) can be used to achieve such
power. Although a hydraulic manipulator can be serial or
parallel linked, it is classified differently here because of the
actuation method and the promise of high torque. A hydraulic
manipulator was proposed by [4] to attain high force/torque
ratio. A hydraulic manipulator is well suited for high payload
capacity aerial systems since a hydraulic system is heavy.
However, it is advantageous over electric systems, since it
is actuated by the flow of oil, which results in a smooth
motion, i.e., fewer oscillations for host platform, and for the
same power output, its inertia is less than other systems.
The proposed hydraulic manipulator was not experimentally
tested, so a physical demonstration of the workability of the
proposed system is yet needed.
4) Cable based manipulators: Cable-based manipulators
are easy to implement, however with cable manipulators,
force can be applied by the pulling motion of the UAV only.
Cable-based manipulators are compliant in nature; however,
there is no dexterity. A simple cable based manipulator lacks
the actuators, that can move the end-effector. Therefore,
UAVs with cables, need to reach the target grasping posi-
tion accurately, hence, precise positioning of UAV will be
required to be able to grasp an object. In one example [64]
the object was attached to the UAV using a string, and
a magnetic gripper. A string provides more compliance as
compared to rigid links, regarding the interaction forces
between the payload and the aerial platform, however, during
the grasping process, it is also susceptible to oscillations,
hindering the attachment of the payload to the gripper.
A Stewart-Gough inspired cable based magnetic gripper
can provide compliance in the vertical axis and reduced
oscillations in the lateral axis. The compliant nature of cable-
based manipulators is very useful when multiple UAVs are
used to lift an object [90], [5], [91]. This allows the UAVs
to space out in the air, preventing chances of collisions. The
orientation of the payload, cannot be altered by a single cable
UAV system, however, in case of multiple UAVs carrying an
object, it is possible to change the orientation of the payload
by changing the relative configuration of the UAVs [90], [5],
Ref DoF Purpose Environment Aerial platform Year
[78] 1 DoF Avian inspired grasping Indoors via Vicon Motion sensors Quadrotor 2013
[72] 9 DoF Hyper-redundant aerial manipulator Indoors Test gantry 2013
[79] 2 DoF Controller design Indoors via Vicon Motion sensors Quadrotor 2013
[36], [16], [35] 7 DoF Grasping via redundant DoFs Outdoors GPS navigation,
vision for manipulation Helicopter 2013-14
[18] Dual 2 DoF arm Valve turning Indoors but via visual servoing Quadrotor 2014
[80] 2 DoF Grasping using model pridictive control Indoors via Vicon Motion sensors Quadrotor 2015
[?] 5 DoF Reduction of disturbance by
keeping actuators close to base NA Quadrotor (intended) 2015
[38] 3 DoF Inspection by contact Simulation and outdoor experiement Octrotor 2015
[21] 6 DoF Battery motion to counteract
manipulator disturbance Indoors via Vicon Motion sensors Octrotor 2015
[81], [82], [83] 3 DoF Open a drawer, grasp a cylinder Indoors vision and vicon Quadrotor/hexrotor 2015-16
[56], [74], [84] 3 DoF
single/dual arm
Compliant interaction and
minimizing disturbance
Indoor/outdoor experiment
with visual servoing Fixed base/hexrotor 2015-17
[85] 3 DoF Self CoG adjusting arm design, cancellation
of torque produced by arm actuators. Indoor and outdoor GPS Hexrotor 2017
[86] 1 DoF Force application (20N) Outdoors Octrotor 2017
Fig. 5. A summary of different methods adopted in decoupled control scheme of UAV and manipulator
Previous sections II and III covered the literature review
regarding the hardware used in aerial manipulation. Current
section will focus on the high-level planning, control, and
environment interaction of the UAVs in subsections IV-A,
IV-B and IV-C respectively.
A. High level planning
High-level planning is crucial in the cases where multiple
UAVs have to work for a shared goal cooperatively. A high-
level planner in these cases plans the mission and then
commands the low-level controller of each UAV. A similar
scenario was presented in [63], [66], [92] where high-level
planning was required to ensure optimization, obstacle avoid-
ance, and co-operation between multiple UAVs. Specifically,
the UAVs were required to autonomously, navigate in the
shared area, detect the colored objects, grasp them and move
towards the drop zone to drop them. Similarly, if the object is
moving, it is crucial to detect the object and its motion, and
then trajectory planning is required for gripping the moving
object [93]. After grasping the object, planning is required
to track the trajectory of grasped payload [94].
Before planning an operation such as grasping an object
via UAVs, it would be useful to extract some information
about the objects in the environment. This information could
be the shape and pose or orientation. Object extraction is
basically a concept where UAVs can get the information
about the payload or the object which is supposed to be lifted
or manipulated. Object extraction could be very useful where
the mission control does not have prior information regarding
the shape and pose or orientation of the payload. This is
very critical to aerial manipulation since grippers require the
compatibility with the shape of the object. Object extraction
relieves the necessity of prior access to the model of the
payload. An onboard object extraction method was developed
by [95] that calculates information necessary for autonomous
grasping of objects. In [96], a CNN (Convolution neural
network) module was used to detect the tree branch where
the UAV can perch. In [97] a Field Programmable Gate Ar-
rays (FPGA) implementable vision-based orientation control
system for an aerial robot is developed, which allows the
multi-rotor to perch on bar like objects. A monocular camera
was used in [97] along with an FPGA. The algorithm which
can be implemented in FPGA is useful specially when high
performing GPUs are not attached or available to the UAV.
After getting information about the object to be manipulated
via object extraction methods or having prior information
about the object, the UAV-manipulator system can interact
with the object.
B. Aerial manipulator system control
The control of a UAV-manipulator system brings forth
challenges such as balancing of ground force effects, forces
by the load, hover precision, flight stability, aerodynamic
disturbances and the effects of manipulator motion. The
motion of manipulator results in the change of inertial
parameters. Addition of multiple links, joints, and actuators
can increase the DoF, but it comes with a price, i.e., increase
in the disturbances related to inertia. In case of multi-
link serial manipulators, the aerial platform‘s control may
require a real-time update of inertial parameters as per the
movement of the arm joints. Another constraint faced by the
UAVs during aerial manipulation, especially in autonomous
load transport is the disruption to flight stability by the
instantaneous addition of a payload or addition of a payload
with offset causing bias forces. Any torque that acts on the
manipulator’s joints, from environment interaction, is trans-
ferred to aerial platforms. A mechanism to deal with such
reactionary moments is also required. The following subsec-
tions will discuss the strategies presented in the literature
which provide solutions to the above-mentioned problems.
The control strategies for an aerial manipulator system can be
divided into two methods. One of the approaches is to have a
centralized controller for both the UAV and the manipulator,
and the other method is a decentralized controller.
1) Centralized control: A centralized controller, for a
UAV-manipulator system, considers the complete dynamic
model of the UAV-manipulator system, and it controls both
the aerial platform and the manipulator. One of the cen-
tralized control was developed by [30]. A combined model
was developed for the helicopter-manipulator system, and
then a trim point based on manipulator mass was used
to linearize the system. A linear quadratic regulator was
developed and tuned to control the system. This approach
has its limitations since it requires the knowledge of trim
points based on the manipulator mass. The performance
of the controller developed by [30] was acceptable in the
steady-state region, and it is not expected to perform well
in disturbances and away from the trim points. The authors,
therefore, proposed the development of a centralized non-
linear control scheme for UAV-manipulator system. Till date,
a centralized non-linear controller is not available in the
literature. Furthermore, the controller developed by [30] was
tested in simulation; hence a hardware experiment is needed
to verify the performance.
2) Decentralized control: A decentralized control strategy
involves two separate controllers for UAV and manipulator
system. In this type of control strategy, the manipulator
motion is considered a disturbance to the control of the
aerial platform. The control of an under-actuated multi-
rotor, with or without a manipulator, requires an outer loop
position control which contains an inner loop attitude control.
When a manipulator is added to the multi-rotor, the attitude
control has to compensate for the manipulator motion or any
reactionary torques. The following subsections are about the
inner-loop attitude and outer loop position controllers for
multi-rotors with manipulators.
Attitude control of UAV-manipulator: A decentralized
control strategy requires some special considerations in terms
of the stability of the whole system. A moving manipulator
causes some instability in the flight of the UAV. This can be
compensated by updating the inertia and CoG of the aerial
manipulator system in real time. As shown in Figure 5, it is
useful for the attitude controller to get a real-time update of
the reactionary torque acting on the manipulator [98], [99],
[100], [101]. The reactionary torque can be obtained by plac-
ing a force sensor on the base of the arm manipulator, or it
can be calculated by using the joint angles if the force sensor
is attached to the end-effector. Another approach to deal with
the instability caused by moving manipulator is to use gain
scheduling, [71], [98] where the gain depends on the real-
time update of the joint angles of the manipulator. In other
approach manipulator’s joint angles can be identified which
can provide the least disturbance during flight [62] for simple
PID controller. Researchers [102], [16], [103], [104] have
suggested, that traditional PID control for the attitude loop
is not sufficient for having a smooth performance. An outer
loop based on Model reference adaptive control (MRAC)
was proposed in [102] to remove the oscillations that are
not prevented by PID control loop only. Other controllers
proposed for attitude control are Direct fuzzy logic [105],
Adaptive fuzzy logic controller [105], Quaternion-based
backstepping approach [103], IDA PBC [104], Variable
parameter integral backstepping controller [106], feedback
linearization approach and finite-time control of the dou-
ble integrator system [99]. Quaternion-based backstepping
approach [103] was able to compensate for the changes
caused by movement of the arm. Variable parameter integral
backstepping controller [106] was able to overcome the
oscillations that were happening while using PID attitude
controller. The double integrator system [99] was able to
reject the parametric uncertainties and external disturbances.
The intelligent controllers [105] were able to provide good
trajectory tracking with the payload attached. IDA PBC [104]
provided bounded trajectories for the bounded motion of the
manipulator. In all the above examples goal was to improve
the attitude control of a UAV with a manipulator, however,
in one study [107] researchers proposed that it is possible
to exploit the motion of manipulator to intentionally change
the CoG of the UAV-manipulator system, so that the limits
of maximum attitude angle can be increased.
Position control of UAV-manipulator: The constraints of
imperfect position control of UAV-manipulator, during aerial
manipulation, can be addressed by sharing the UAV‘s po-
sition or position error information with the manipulator
controller, so that manipulator controller can compensate for
the UAVs position errors by moving the manipulator accord-
ingly [106], [103], [34]. Similarly, end-effector‘s position or
position error can be shared with the position controller of
the UAV, so that position controller can compensate for it, by
moving the vehicle accordingly [106], [103]. This strategy
is helpful, in cases, where decentralized control of UAV and
manipulator is achieved by considering the movement of
the manipulator as perturbations to the host platform. One
approach is to use adaptive outer-loop horizontal position
and altitude control [108]. The adaptive altitude controller
in [108] generates the required thrust to hover, after con-
sidering the error in altitude, while adapting to any change
in inertia and mass due to the motion of the manipulator.
Similarly, the adaptive horizontal position controller, gen-
erates the attitude commands after considering the error in
horizontal position, while adapting to any change of inertia
and mass due to manipulator motion. The resulting adaptive
position controller was able to stabilize the platform, from
the disturbances caused by the movement of the manipulator.
Aerial manipulation, especially, for near ground opera-
tions such as picking up an object, would be prone to
the disturbances caused by aerodynamic forces arising due
to proximity to the ground [109]. The solution proposed
by [109] was to quickly descend after spotting the object
and lifting it up. One of the solutions for this limitation is to
let the UAV hover above the object to let the UAV settle, and
when the error in position is at its minimum, the next set-
point for the position can be sent to start the descent towards
the object.
C. Environment interaction
The control of the aerial platform with a manipulator
was discussed in the previous section IV-B, which allows
the UAV-manipulator system to be able to move in the air
while staying stable. However, an important requirement for
the usefulness of a UAV-manipulator system is its ability
to interact with the environment. This interaction could be
performed via, different approaches as shown in Figure 6.
The following sections will discuss the impedance control,
force control, torque control and object interaction via visual
1) End-Effector interaction control: The ability of the
UAV-multi-DoF-manipulator system to have a stable hover
with moving manipulator, i.e., varying joint angles is crucial
for aerial manipulation. Moreover, another advanced aspect
of UAV-manipulation is its end-effector‘s stable interaction
with the environment. The stable interaction of the manipula-
tor‘s end-effector can be achieved by making it compliant to
external forces. It was discussed in section III-B.1 that com-
pliant behavior can be achieved by using elastic elements, but
elastic elements have fixed behavior, however, if compliance
is achieved by using impedance control, the elastic and
damping, behavior of the system can be configured based
on the requirements [50], [110], [111], [112], [113], [71],
[114], [115]. The impedance control can be classified based
on the final control signal, which could be the joint angles
of the arms or position based impedance control, or torque
acting on arm joints. Apart from this basic classification, the
sequence based interaction control structure is described in
Figure 7.
1.1) Position based interaction control: The preference
of position based interaction control over torque based in-
teraction control comes from the fact that [71], although
some larger servo motors may have the capability to control
torque, the choice of joint angle based admittance control
is due to poor accuracy of torque sensors. It is shown
in Figure 7 that position based impedance controller is
implemented in literature consisting of multiple layers [112],
[111], [50], whereas a position based admittance controller is
implemented in [71]. In case of [112], [111], [50], the end-
effector‘s reference trajectory is provided to the impedance
Fig. 6. A single UAV-manipulator system‘s interaction control hierarchy
controller, while in case of [116] joint angles are provided to
the impedance controller. The required joint angles are sent
to the joint controller in the end for position based impedance
Impedance control [50] takes the required or desired tra-
jectory and based on force sensor input data and the current
trajectory, the impedance controller refines the trajectory, to
reduce the effects of reactionary forces. This new trajectory
is then implemented by a motion controller. An admittance
controller was proposed by [71], which calculates the desired
position of the end effector, as the sum of the distance from
the desired position and the distance that is required to exert
the necessary force. Another similar approach was presented
by [112] in which controller was based on an impedance
outer layer, which generates a compliant trajectory that
results in bounded forces of interaction with the environment.
The inner loop uses inverse kinematics to generate joint
commands, while the innermost loop tracks the motion.
In case of [116], a motion planner, on the basis of the
desired trajectory for the end-effector, determines the desired
motion for the actuated variables (joint positions, quad-
rotor position, and yaw angle). The second layer utilizes an
impedance filter to provide a selective compliant behavior to
the system, for the interaction of the manipulator end-effector
with the environment. Finally, the third layer implements a
motion controller aimed at tracking the references output by
the previous layer.
The multi-UAV grasp of an object with multi-link serial
manipulators presents a challenge for control of the inter-
action forces between UAV-manipulators and the jointly-
held object. The object or payload, if grasped by two
UAV-manipulators can have any unbounded force flow from
one UAV-manipulator to the other UAV-manipulator. This
unbounded force flow can de-stabilize the flight. The un-
bounded force flow can be bounded by an impedance con-
troller as proposed by [111]. First impedance layer generates
the reference trajectory to limit the external forces while
the second filter deals with the internal forces between the
manipulator and the UAV. The trajectory generated by the
two filters are then used by a motion controller which uses
inverse kinematics and a PD controller to implement those
reference trajectories. Another approach for aerial manipu-
lation with multi-link arms is discussed by [110], where a
three-layer control architecture is proposed. The first layer
is a centralized layer that generates trajectories for all end-
effectors. The second layer is local to all UAV-manipulators,
which computes motion references to track such end-effector
trajectories coming from the centralized upper layer. The
third layer is a low-level motion controller that tracks these
motion references. It seems that a combination of approaches
by [111], [110] would be better than using them individually,
as explained below. Addition of the high-level top layer
of [111] to the control strategy of [110], would make the
process even safer and smoother. Both [111], [110] are
about multi-UAV manipulation, but in [111] the internal
forces between manipulator and UAV are also accounted,
and in [110] a high-level layer takes into consideration of all
end-effector poses.
1.2) Torque based impedance control: A different ap-
proach is torque based impedance control proposed by [113],
in which the control input is the desired trajectory and current
trajectory, and the control signal is the torque commands sent
to the arm joints. Implementation of impedance control using
torque would require taking care of sensor inaccuracies.
The controller should be made [113] robust against any
bounded force sensor inaccuracies and bounded unstructured
modeling (non-parametric) uncertainties and/or disturbances
in the system. In [117] similar strategy was presented, but an
adaptive controller was proposed that estimates the dynamic
model parameters.
2) Contact force control: Applications that require con-
tact between an aerial manipulator‘s end effector and the wall
would require the aerial platform to be able to control these
forces. This type of force control is effective in contact based
sensing applications, where a smaller delta manipulator with
a sensor is attached on the side of a UAV. It was shown
by [19] that when a UAV with a delta manipulator on its side,
comes in contact with a wall, due to the reference position
being virtually inside the wall, the actuators of the multi-
rotor, will exert a force similar to a spring mass damper
system. The force exerted during contact with the wall can
be regulated by careful selection of the lateral reference
position of UAV and manipulator set-points. A PI force
controller was implemented by [89] on the manipulator in
the direction normal to contact to regulate the force applied.
A UAV-manipulator system was also developed by [118]
for application of force, in which after establishing contact,
desired attitude angles are calculated based on the required
force; however, there was no feedback involved regarding
the estimation of force or sensing of force for its regulation.
In [86] a 1 DoF arm was used to achieve a force of 20N
for the hammering inspection of old walls. In this study,
the force sensor was used, and force control was integrated
with position control. The mean error was around 4.27 N
which might suit the hammering test but not suitable for
other applications where more accurate force control will be
3) Applying torque control: Yaw motion of a multi-rotor
can be used to apply torque [7]. The yaw motion of a multi-
rotor is caused by the difference in the angular speeds of
the counter and clockwise rotations of the propellers. The
torque generated by the multi-rotor is theoretically the sum
of reaction torque produced by each propeller of the multi-
rotor. An impactive gripper as shown in Figure 3(d), just
attached below a multi-rotor was used in [7] to demonstrate
unscrewing a light bulb. An increased amount of torque
can be applied, via using dual arms attached to a multi-
rotor [18], by increasing the moment arm between the forces.
The torque applied can be increased further, if torque is
applied by multiple UAVs. A similar concept was presented
in a newly developed SmQ platform by [47], where multiple
quadrotors were attached to a single frame, and a yaw motion
was demonstrated. This concept can be exploited to achieve
a high torque and would be useful in applications where high
torque is necessary. It is also possible to not only control the
torque by increasing and decreasing the force, but also by
changing the point of application of the force. The point of
application of force can be changed by simply attaching the
multiple UAVs, closer or farther on the object.
4) Environment interaction via visual servoing: All the
above control strategies involve the position based or torque
based control methods, but recent studies by [82], [119],
[120], [51], [121], [20], [64], [35], [122] utilized visual ser-
voing to perform aerial manipulation tasks. Visual servoing
requires attachment of the camera, which adds a little more
mass to the UAV, and needs a bit of extra power to use the
camera. Visual servoing is also very dependent on lighting
conditions, and it won’t work with insufficient lighting. This
section describes different uses of visual servoing technique
in aerial manipulation. The tasks including navigating to-
wards payload, directing the host platform towards the de-
sired location, maintaining a ready pose of the manipulator,
picking up the object, transporting the carried object towards
the target location, are performed in studies described below
using visual servoing. Apart from visually servoing the aerial
manipulator system towards the payload, [64] showed vision
based search, docking, and lifting of the payload. The authors
used visual servoing to approach and attach the magnetic
end-effector to the correct spots and then lift the object up.
The next subsection IV-C.4.1 describes the types of visual
servoing methods.
4.1) Types of visual servoing: Visual servoing can be done
in two ways, position based visual servoing [120] and Image-
based visual servoing (IBVS) [51], [20]. In position based
visual servoing, markers are used to estimate the position,
and then a control strategy is developed based on position
estimation, however, an IBVS is based on control law, which
is based on the error between desired and current features
on the image plane, and does not require any estimate of the
position. Position based visual servoing is least susceptible
to lead the marker outside of the field of view, whereas
image-based visual servo (IBVS) is usually done by velocity
commands and there is always a chance of losing the sight
of the marker. Researchers suggested using fish eye [82]
camera to enhance the field of view for the case of image-
based visual servo, which then requires to do some image
processing to convert it into perspective camera image.
During roll and pitch the platform tilts; therefore the image
is distorted, an image adjustment method is developed to
address this issue. The control strategy starts with the raw
image, which is used to calculate blob image. The blob image
is converted from fish eye to perspective camera image. After
extracting contour, the compensation of roll and pitch are
performed, and then moments are calculated from the image
to be used to calculate the velocity commands [82]. Authors
including [119], [82], [51], [20], [64] developed their control
strategies based solely on IBVS whereas [120] combined
both image-based visual servoing and position based visual
servoing by means of selecting one of the above methods
based on the task. The tasks are defined as first moving
towards the object and maintaining the desired position,
which is achieved by position based visual servoing with
the help of a downward facing camera. The next task is the
end effector positioning and orientation which is achieved
by image-based approach.
4.2) Camera configurations: There are usually two types
of camera configuration found in literature, eye-to-hand, and
eye-in-hand. Eye-in-hand configuration can support visual
servoing of the hand‘s end effector towards the target po-
sition, while eye-to-Hand can not only provide information
about the hand position with respect to the target position but
eye-to-Hand can also tell us the relative position of the hand
with respect to the platform. Several approaches require the
use of eye-to-hand configuration such as Self visual servoing
(SVS), will be discussed in the eye-to-hand section below.
Eye-in-hand: In eye-in-hand, the pose of the camera varies
with the motion of the robotic arm as well as the multi-rotor.
Image-based visual servoing with eye-in-hand camera config-
uration for aerial manipulation was used by [82]. The study
is focused on combined model of aerial manipulator, and
a passivity-based adaptive controller is designed which can
work on both position and velocity control. This controller
is applied to position control so that the aerial manipulator
can reach the location of desired objects. Velocity control is
used with IBVS to guide the aerial manipulator system.
The papers [51], [20] implement the IBVS using the
eye-in-hand camera configuration. The authors use the con-
cept of a ready pose of manipulator with respect to the
target object, which enables the manipulator to quickly
move around the target object as desired. The goal of the
manipulator controller is to maintain that ready pose by
using IBVS while considering the host platform‘s motion
as perturbations. Simultaneously the kinematic information
obtained from manipulator is used to find the difference
Fig. 7. Different sequences of End-effector‘s interaction control with environment as proposed in literature
between the ready pose and arm‘s current pose, and it is
used to direct the host platform to move. Eye-in-hand camera
configuration was used by [71] for relative positioning of
end-effector with respect to the object of interest. Position
based visual servoing was used, markers were placed on
the object to identify object‘s relative pose and orientation
with respect to the end-effector. In eye-in-hand approach,
however, after grasping an object, there is a chance of losing
some field of view, therefore after grasping the object, if
visual servoing is to be used, an eye-to-hand configuration
would be well suited. In [123] a dual arm aerial manipulator
is proposed, where one arm is equipped with eye-in-hand
camera configuration while the other arm is used to perform
the interaction with the environment. The visual information
is used to regulate eye-in-hand camera motion with respect
to the arm performing the task. The work [123] however
needs experimental testing.
Eye-to-hand: Eye-to-hand camera configuration was used
by [119] for the placement of the payload on its target
location. A control law was developed which calculates an
error via IBVS to generate the velocity commands for both
UAV and manipulator [119]. The error is based on the image
error that is obtained using an eye-to-hand configuration
camera. The resulting robotic operation is thus reduced to
automatically position an assembly part to the desired loca-
tion. The placement of payload on target location was also
demonstrated by [121] who proposed simultaneous control
of UAV and manipulator via velocity commands generated
by using image-based visual servoing with the help of an
onboard eye-to-hand camera. However, what differs in this
study from [119] is the new IBVS technique called self visual
servoing (SVS). In this case, the error is the difference in
superposition of image features in the carried object and the
target‘s image. Eye-to-hand configuration does not provide
the relative position of end-effector with respect to object,
but it can be obtained by the knowledge of joint angles or by
placing a marker on the end-effector itself. This increases the
computational load. There is also a possibility of occlusion of
the target object by the manipulator or the end-effector. Eye-
to-hand configuration was also used by [122] which was used
to get the pose of the object with known markers. In [122]
the camera was attached to a 1DoF joint which allows better
tracking of the object.
Latency requirements: The positioning error scales di-
rectly with the larger aerial manipulator systems; the end
effector position error linearly scales up with the radial length
of the manipulator and end effector because of the error
in yaw of the aerial platform. It was shown by [35] that
latency requirements are crucial for this type of system and
time delays in propagation of the signal between perception
and actuation components which can significantly affect
the overall performance of a visual servoing system. This
constraint can be solved by calculating the time delay and
to counteract the time delay based on the predicted motion
of end effector. The method was verified using experiments.
Development of aerial platforms for transportation of
heavy payload poses restrictions regarding scaling up the
UAVs such as multi-rotors as discussed in subsection II-
A, so it is intuitive to use multiple UAVs to transport a
heavy payload. Payloads with complex shapes are also easier
to carry using multiple UAVs. The orientation control of a
larger payload is difficult to manage with a single UAV yaw
motion. Multi UAV-collaborative-aerial-manipulator system
should be able to deal with the forces imparted by the other
UAVs, that are coupled together physically. If an object or
payload is held by multiple UAVs, the goal of the control
system is to avoid collision amongst UAVs and to ensure
that the payload or object is following the desired trajectory,
and also the forces acting on the object by the UAVs are
bounded [111], [124]. The challenges associated with multi-
UAV collaborative manipulation also include the difficulty
in collision avoidance path planning since the volume of
the moving system significantly increases with multi-UAV
aerial manipulation [125]. On the other hand, having multiple
UAVs provides an opportunity for cooperative localization
which can be done via sharing the position estimations of
each vehicle with the other [126]. Another challenge is
the payload parameter estimation. The inertia and mass of
the payload are assumed to be known for the collaborative
transport. In [127] a method to estimate inertial parameters
of a payload of known geometry is provided. However, in
a factory setting or unknown environments geometry of the
payload might not be known from before. The control of
physically coupled systems can be either via co-ordinated
motion or another approach of leader-follower. The following
sections V-A and V-B will discuss the leader-follower and
co-ordinated motion method in detail.
A. Leader follower approach
The leader-follower method requires one of the aerial
platforms to take the lead, and the follower aerial platform
adjusts its motion based on the motion of the leader. Leader-
follower approach for multi-UAV transportation is useful
for the cases where the communication between the two
aerial platforms is susceptible to communication outages.
It does not require a centralized or ground control station
to command each aerial platform. However, the challenges
associated with the leader-follower approach is the ability to
perform complex maneuvers and full attitude control of the
payload. Leader-follower approach without communication
between two UAVs can be done via two methods. In one of
them, follower UAV depends on visual cues from the leader
UAV and the gripper and in the other follower UAV applies
passive force control to adjust to the motion of the leader.
The discussion regarding these two approaches is provided
1) Leader follower approach using visual cues: One ex-
ample of the leader-follower approach is [24] where authors
used visual cues for collaboration between leader and fol-
lower. One marker is placed on the Leader UAV which allows
the follower UAV to maintain altitude and heading alignment
with the leader. A pendulum based 1 Degree of freedom
manipulator is attached to each UAV with magnetic grippers.
Another marker is present on each side of the magnetic
gripper‘s attachment point of the cylindrical payload. The
follower keeps itself on top of the payload by tracking the
markers on payload using a force controller(PD) based on
the distance error, i.e., the offset of the marker from the line
below the UAV. In this case only forward or backward motion
is allowed, but any lateral movement is not considered.
2) Leader follower approach using passive force control:
Another recent example of the leader-follower approach
is [128], where the authors use passive force control for
the follower UAV while the leader is following the required
trajectory. Passive force control was achieved by using an
admittance controller that modifies the reference trajectory
for the follower by calculating the force acting on the
follower. The admittance controller can be tuned by changing
the stiffness of the assumed spring mass damper system.
More stiffness results in a strict following of the initial
trajectory, whereas low stiffness means better compliance to
the external forces. The external force acting on the follower
is calculated by using the state estimates of the follower
and the rotor speeds. There is a possibility of incorrect
estimation of force when the follower UAV might perceive
any uncertain external force such as wind gusts as the force
exerted by the leader. Passive force control seems restricted
in orienting the payload in only certain scenarios. Another
study by [129] however addresses this problem by choosing
a non-zero internal force to induce a required attitude of the
payload. This method was extensively simulated by [129] but
experimental tests will be required for further assessment of
the method. In passive force control, the leader can also not
change the direction of motion of the system suddenly or
move towards the direction of the follower, while in case
of [24], the leader may even move towards the direction of
follower and follower may adjust its position. However, these
methods are unlikely to be able to perform maneuvers to
follow slalom path as it can be performed in co-ordinated
motion [126]. The study [128] assumes that the leader
and follower altitude are aligned properly, and the method
relies on state estimates that were provided using a visual
inertial navigation system built for the indoor and outdoor
B. Co-ordinated motion
Apart from the leader-follower approach, another method
of multi-UAV manipulation is via co-ordinated motion. Co-
ordinated motion is defined as a multi-UAV transport strat-
egy, where a centralized station is required for information
exchange between the multiple UAVs. In some cases, the
centralized station acts in the form of a centralized trajectory
generator and controller. Co-ordinated motion method could
be in the form of rigid physical coupling between the UAVs
and the held object, or it could be in a relatively flexible
physical coupling such as holding the payload using cables or
multi-link arms. In the case of flexible systems, the scenario
can be compared to that of formation control.
1) Formation: Several studies related with the multi-
UAV collaborative transport of an object showed reliance
on maintaining certain configuration in the air, which is
a formation approach. These approaches for collaborative
transportation are based on a centralized solution, where a
centralized system computes a control action for each of
the UAVs, and it shares the command with them. Most of
these studies are based on the absolute position estimation
of the UAVs and not the relative positions. Manipulating an
object using formation, or a configuration of UAVs in the
air is usually achieved by using cable-based systems. The
advantage of using cables instead of manipulators or grippers
in these systems is that the extension of cable provides more
space to the UAVs to spread-out while in flight and thus
reduce the chances of collision during formation flight.
An initial study by [31] based on multi-UAV manipulation
by co-ordinated motion showed the multi-UAV payload
transportation by using three helicopters and cables in out-
door environments. However, a deep analysis was required
to ensure static equilibrium of the payload and maintain the
stability of the system. Therefore, studies [90], [5], [91] are
mainly concerned with maintaining the equilibrium of the
payload. In general, the orientation of the payload and the
trajectory depends on the position control of the UAVs. The
orientation or attitude control of the payload can be useful,
in the case of narrow passages, where a particular orientation
could be required to pass through.
Specifically, [5] developed spatial configurations for the
UAVs that resulted in the static equilibrium of the payload.
These configurations were developed considering the con-
straints on the tension of the cables. These configurations
enabled controlling of multiple UAVs while transporting or
manipulating the payload. The proposed solution by [5]
may result in multiple payload equilibrium solutions. So,
non-trivial solutions can be obtained [90] by developing
constraints for the UAV spatial configuration that guarantee
the existence of a non-trivial payload pose. Another approach
regarding the static equilibrium of the payload was presented
in [91], where the solution of UAV spatial configuration, was
found using inverse kinematics and equilibrium conditions.
The authors claimed that if cable tensions are specified, there
exists a finite number of solutions and an efficient analytic
algorithm based on dialytic elimination was used to find
those solutions for the spatial configurations of the UAVs.
A rather different approach was from [130] where model-
ing conventions of reconfigurable cable-driven parallel robots
(RCDPR) was used to derive direct relations between the
motion of quadrotors and the motion of payload. This method
does not require specification of tensions in cables and uses a
tension distribution algorithm to distribute the cable tensions
The formation approach requires some considerations
based on payload type, in all the above cases, the payload is
a rigid body with a certain width. There are some cases when
the payload is similar to a point mass with a certain width or
when it is a deformable linear object (DLO) [131], such as a
hose, or cables for building rope-bridges, etc. In such cases,
when the DLO is more massive, it would require multiple
UAVs to transport it. The weight distribution, in this case,
is unequal if all the UAVs have the same altitude, so robot
configurations are found using particle swarm optimization
that can provide equal-load distribution amongst multiple
UAVs carrying the payload [131]. In case, when payload
shape is similar to a point mass, i.e., with a smaller width,
there are certain implications. During the lift, the UAVs
will experience a pull towards each other, and during the
transport, this force must be taken into account [132].
The cable-based systems pull the payload, usually upward,
but in case, where force application is required a concept
presented by [133], [134] could be useful. This concept
is multi-UAV collaborative manipulation system based on
swarm robotics. The proposed system called, the flying hand
is a robotic hand consisting of a swarm of UAVs able to grasp
an object where each UAV contributes to the grasping task
with a single contact point at the tool-tip. The flying hand,
in [133] required the human in the loop, and the UAVs were
following the hand gesture of the human. In [134], however,
haptic feedback was used. These studies are different because
the UAVs collaborate to become a grasping mechanism,
while in most cases a gripper with or without an arm
manipulator is utilized.
2) Rigid coupling: A payload which is grasped by mul-
tiple UAVs using grippers which are rigidly coupled with
UAVs does not provide the compliance that is available in
case of cable-based multi-UAV and multi-link-manipulator
based transportation. The rigid coupling while works for
transportation, but it restricts the ability to change pose
and orientation of the payload as compared to cable-based
systems. The absence of multi-link arms also restricts the
ability to move the payload in 3D space while keeping the
UAVs in hover. The advantage of having rigid coupling is
the availability of constraints in space and time between
consecutive position estimates [126] provided by each UAV.
If the payload is rigid and the geometry is known, having
rigid coupling can, therefore, assist in position estimation in
GPS denied environments. In this arrangement co-operative
localization is performed where each UAV can benefit from
the measurement performed by other UAVs. One study [61]
uses grippers attached to the base of the quad-rotor. In this
case, a control law is developed based on the payload shape.
The proposed control law is based on controlling individual
UAVs on their grasping point on the payload. The agents
know their grasping position on the payload and the common
goal. Due to the knowledge of the grasping position, the
orientation and angular estimation of the individual UAV is
used to calculate the orientation and angular velocity of the
payload. The position and velocity of the center of mass are
calculated from the position and velocity of individual UAVs.
Each UAV then runs a local hover or velocity controller
along with the attitude. In this type of collaborative transport,
which relies on centralized co-ordinating station smooth
communication is important, any outage in communication
can destabilize the system.
Indoor experiments use motion detectors such as Vicon
motion sensors [135] or OptiTrack motion capture systems
etc., for accurate tracking of reflective markers indoors.
These markers could be placed on objects in order to track
them with high precision indoors. Outdoor experiments,
however, suffer from localization inaccuracies due to the
limited GPS accuracy and update rate. Moreover, urban envi-
ronments with high rise buildings are particularly challenging
due to the fact that these environments can block satellite
signals and introduce multipath signal challenges that further
reduces the accuracy of the location. Localization accuracy is
particularly important in aerial manipulation since the UAV is
expected to physically interact with the objects or surface. In
multi-UAV collaborative aerial manipulation, the localization
inaccuracies may result in undesired forces on the jointly
carried payload [111].
RTK has been used in certain environments [71] to en-
hance the quality of the GPS signal and provide a centimeter
level localization accuracy. This, however, requires additional
RTK base station to be placed in the region where the
aerial manipulation is performed. Apart from RTK GPS,
techniques such as the Visual Simultaneous Localization
and Mapping(VSLAM) [109] has been used in GPS denied
environments to handle the outdoor localization challenges.
Vision-based navigation [109] , visual servoing can be used
for outdoor tasks to handle positioning inaccuracies and
provide an accurate local relative position suitable for aerial
manipulation [92].
Outdoor aerial manipulation should also handle external
disturbances introduced by environmental factors such as
wind gusts. Disturbance rejection controllers such as [136],
[137], have been developed in the past to handle such
disturbances. In case of multi-UAV collaborative transport,
the outdoor environment can suffer from communication
outages, position inaccuracies, and in-sufficient lightning
conditions for vision-based methods.
The energy consumption in multi-rotors is critical since
they are powered by batteries which have limited capacity.
The major power consumption results from the motors which
are rotating the propellers to generate thrust that keeps the
UAV airborne. The electrical energy consumed by the motors
depend on the thrust requirements, and also includes the
electrical losses and overall propulsion system efficiency.
These electrical losses include losses in motors, losses in
electronic speed controllers. The resulting flight time of
multi-rotors is around 20 to 30 minutes [26]. Attaching a
moving manipulator also adds to the energy budget of the
aerial platform, regarding added weight and also the energy
required by the actuators of the manipulator. So far no study
is found which addresses additional energy requirement due
to the addition of the moving manipulator or a gripper
with actuators. Other causes of energy consumption include
the autopilot or any companion computer attached to the
aerial platform, sensors such as camera for visual servoing
or communication links. The energy constraints in aerial
manipulation can be addressed in various stages. One stage is
the design stage of the aerial manipulator. The second stage
is the energy savings by efficient planning of the operation.
The energy savings by design of the aerial manipulator and
by efficient planning are discussed below.
Energy savings in design stage can be done in various
ways, including reducing the amount of weight carried by
the UAV. In some cases, the aerial platform, relies on
support from the surroundings, in order to save energy to
keep hovering, one example is a novel mechanism proposed
by [43] which relies on anchors to keep the platform sup-
ported, instead of using the thrust of the propellers. Similarly
a gripper on top of the multi-rotor can provide support
while the attached manipulator can perform the required
operation [138]. In case of [3] the drone supports itself on the
wire it is inspecting and hovers only to reach and dock on the
wire, or to avoid obstacles during its travel on the wire. If the
aerial platform is based on a hybrid design, such as [42] can
travel on ground when flight is not necessary to save energy.
The aerial platform can perch while doing manipulation to
avoid continuously generating the thrust [139]. The energy
saving in the second stage is done via efficient planning. In
these cases, based on the motor dynamics, minimum energy
consumption path can be generated for a single UAV [26].
It is also possible to chose minimum energy consumption
path from multiple available paths, generated by a path
planner [48] for a single UAV. Another important factor is
the mass of the UAV, [140] showed that there is an optimum
mass which results in maximum endurance of the UAV
flight. In case of specific applications where close contact is
required by the UAV, ceiling effect can be used to maximize
the flight time [8]. The ceiling effect is similar to ground
effects when a UAV is approaching a roof like surface from
below it induces additional thrust.
Energy management in multi-UAV collaborative transport
is important since power failure in one of the UAVs can
fail the whole operation. Although several demonstrations of
multi-UAV collaborative transport are present in literature
as discussed in section V, none of the papers addressed
the energy distribution problem amongst the collaborating
UAVs. A study by [141] discussed the need of equal load
distribution, in case of transportation of deformable linear ob-
jects, the configuration of multiple UAVs which can provide
equal-load distribution was estimated using particle swarm
optimization. However, in that case, it is assumed that all
UAVs are similar and their battery capacities are also the
same. Several papers as discussed in section V are also
concerned with changing the pose and orientation of the
payload in air [5], [91], while not considering the un-even
thrust requirements and hence power distribution. The study
by [142] considered the development of a multi-objective
control strategy for multi-UAV collaborative transportation
where thrust values can be assigned to each vehicle. The
mechanism proposed by [142] regulates the thrust require-
ments; however, this regulation should be done considering
the energy availability or the heterogeneity of the UAVs.
An optimization algorithm is required to ensure mission
completion for heterogeneous collaborative UAVs or homo-
geneous UAVs with different energy levels. Furthermore, an
increase in the volume of the payload nullifies the assumption
of point mass and any collaborative transport mechanism
should consider the center of gravity of the payload while
distributing the thrust requirements. Most of the studies
for collaborative transport use objects of smaller width and
larger length (i.e., higher aspect ratios) whereas in case of
lower aspect ratio payload any difference in altitude of the
collaborating UAVs will create an uneven thrust requirement
thus causing uneven energy distribution. An increase in the
depth of the jointly carried payload can also cause shielding
of air flow generated by the propellers of the aerial platform
which can cause reduction in the thrust generated. This will,
in turn, result in more efforts by the propulsion system and
hence more energy consumption. The aerodynamic behavior
of jointly carried object is an open area of research.
In general, the interest of researchers is continuously grow-
ing in this field. Below are some generalized observations
and research gaps which are identified during the survey.
A. Aerial platform
1) The novel VTOL aerial platforms as discussed in sec-
tion II-C which have potential for aerial manipulation
are yet to be investigated for addition of an aerial
manipulator. For example, there is no study regarding
a manipulator attached to a tilt-rotor. Other platforms
such as ODAR, or SmQ, can be tested after attaching
a manipulator.
2) A stability and efficiency analysis is required for
comparing similar scale (capacity and size) helicopters
and multi-rotors and attached with similar scale manip-
ulators. This sort of stability analysis is already done
for addition of payload mass, but it should be extended
to the addition of manipulator as well. This study could
answer regarding the favorable platform for attaching
a manipulator.
3) The transformable aerial platforms although provide an
opportunity for aerial manipulation in terms of variable
shape. Having variable shape can allow an aerial
platform to navigate through tight spaces. However,
this also means that the aerodynamic behavior of the
platform would also vary with the transformations.
The parasitic drag which becomes significant at higher
speeds would also be uncertain, and the problem will
become more complex to deal with.
4) The crash-worthiness of aerial manipulation systems
needs to be studied and further enhanced.
B. Aerial manipulators and grippers
1) To the best of our knowledge, soft grippers and ma-
nipulators are not yet explored with regards to the
aerial manipulation. Soft grippers can provide promis-
ing gripping performance and safe interaction with
humans. Soft grippers should also be explored for their
compliant behavior which can assist in manipulating
various types of objects of irregular shapes.
2) A recent study showed that drone delivery could help
reduce the greenhouse gases [12]. This, however, can-
not be yet said about aerial manipulators. A detailed
study is required to perform a life cycle analysis of
each of those aerial manipulators specific for each
3) Apart from the carbon footprint of the aerial ma-
nipulators, it is also necessary to study the financial
and feasibility analysis of aerial manipulators vs. the
human workers. This is necessary to persuade the
end users of aerial manipulation technology to replace
human workers from hostile environments.
4) Variable stiffness in compliant grippers and manipula-
tors should be explored.
5) Although haptic feedback for grasping is studied [143].
However, haptic feedback for single and multi-UAV
torque applications is needed, since required torque
changes from static to dynamic while opening a valve.
6) Aerodynamic effects of a manipulator mounted on
UAV can be countered while the manipulator is folded,
however, if the manipulator happens to pick an object
and performs translational maneuvers, it might still
result in loses due to aerodynamic drags.
7) Scaling laws can be developed for aerial manipulators
to give a quick idea of how they can be scaled in size,
capacity. This can also be supplemented by a stability
8) There is a ”global conditioning index” [88] about the
measurement of dexterity, but apart from that, other
aspects of aerial manipulation, such as how much
manipulation can be performed while being airborne
is not discussed. Other performance measures such as
financial indicators are not discussed in the literature.
9) A hybrid serial and parallel linked manipulator, could
also be investigated, for its behavior as an aerial
manipulator. A hybrid manipulator can combine the
benefits of both serial and parallel manipulators.
10) In vision-based pick and place of an object, it is
possible that the gripper might hinder the view and
thus lose track of the Tag/marker. A strategy to deal
with such hindrance is needed to ensure smooth pick
and place of the payload using vision.
C. Energy conservation
1) The additional energy requirements for manipulator
motion will reduce the flight time. Although, several
papers addressed the control and design of aerial
manipulators, yet there is no discussion about in-
creased energy requirements in case of addition of
a manipulator to the UAV. A quantification of and
scaling of increased energy requirements would help
investigate the possibilities of commercialization of
aerial manipulation systems.
2) Hardware and algorithm based optimization of the
aerial manipulator to reduce energy consumption
should be performed.
3) One constraint in commercialized aerial manipulation
is the limited energy quota per charge and the time
to charge the battery. Energy constraint brings forth
the payload capacity limitations, which also restricts
the weight of the manipulator. A heavy manipulator
results in lower flight time.
D. Multi-UAV aerial manipulation
1) In multi-UAV collaborative transport, the air flow by
the rotors seems to be shielded by the payload. This
shielding effect appears to reduce the thrust generated
by the rotors. Aerodynamic analysis of such shielding
and quantification or scaling of such shielding effect
is necessary. There is also a possibility of a multi-
rotor to effect another multi-rotor aerodynamically
while jointly transporting a payload. Therefore, a lower
bound on the allowable proximity of multi-rotors while
picking up and transporting a payload in the air should
also be scaled.
2) Multi-UAV transport systems need a mechanism to
decide how many multi-rotors would be required to
pick up a particular object. This decision does not
only depend on the weight of the payload but also
on the shape of the payload. Complex shaped objects
might result in unusually directed torque on the aerial
platforms. Usually in the studies surveyed the problem
of object pickup is either ignored or simplified by
putting markers. If the shape is complex, then pick
up points need to be determined.
3) Multi-UAV collaborative transport should be extended
from DLOs to deformable multi-dimensional objects.
The research in this area will extend the use of multi-
rotors for applications such as using fishing nets in the
4) A multi-UAV system for application of torque needs
further investigation since with a multi-UAV system
we can exert more torque and hence can perform
more torsion related operations. A similar system is
developed as an SmQ platform, which can apply higher
torque using multiple UAVs, but further quantitative
analysis of such a system are still needed, and a
demonstration of the application of torque is needed.
5) Energy optimization in multi-UAV collaborative trans-
port systems and energy distribution amongst the UAVs
during collaborative transport needs exploration.
6) Multi-UAV collaborative transport based on leader-
follower approach, still needs an energy efficient strat-
egy, to transport an object.
7) Multi-UAV collaborative transportation systems are
vulnerable to the failures more than single UAV trans-
portation. Mechanisms to deal with a single UAV rotor
failure are present which can prevent crashing of the
UAV when one or more rotors fail. Similar studies are
needed for multi-UAV collaborative systems.
Aerial manipulation has the potential for application in ar-
eas spreading from agriculture, industry, e-commerce, emer-
gency response, search and rescue. This literature review pro-
vided the state-of-the-art in the field of aerial manipulation
which could benefit the professionals from diverse fields,
to be able to implement aerial manipulation based solutions
to their application areas. This literature survey started with
the state of the art in platform development for aerial
manipulation. Conventional aerial manipulation platforms
such as helicopters and multi-rotors were discussed, followed
by the survey of development of novel aerial platforms.
It was found that platform development is a very active
area of research. After the discussion of aerial platforms,
the literature review moved forward towards classification
of different manipulators and grippers found in literature.
It was found that each class of grippers and manipulators
exhibits certain attributes which were tabulated and discussed
in detail with examples. These attributes include the level of
disturbance to the aerial platform caused by different types of
manipulators. Researchers have attempted to develop designs
which can minimize the disturbance caused by a moving
manipulator, but this area of research is still active, and
it is expected that many more attempts would be made in
recent future in this regard. An important aspect of aerial
manipulation is the safe and stable interaction of the aerial
manipulator with the environment. These interactions could
be for application of force, torque, etc. Researchers have
used impedance controllers to achieve safe interaction of end-
effector with the environment. Visual servoing is another im-
portant means of environment interaction. Researchers have
demonstrated the use of visual servoing to move towards the
object, pick it up and transport it to the target location.
The recent trend of using multiple UAVs for aerial ma-
nipulation was also discussed, with a focus on two basic
strategies, leader-follower, and coordinated motion. All these
approaches, are susceptible to communication outages, and
uncertainties of outdoor conditions, such as insufficient light-
ing conditions or wind gust, etc. The progress made by
researchers cannot be commercialized unless energy con-
straints are resolved. Currently, energy constraints are the
biggest bottleneck in aerial manipulation.
[1] Siciliano b. ijars video series: Aerial manipulation. [Online].
Available: DOI: 10.5772/60874
[2] J. J. Acevedo, B. C. Arrue, I. Maza, and A. Ollero, “Cooperative
large area surveillance with a team of aerial mobile robots for
long endurance missions,” Journal of Intelligent & Robotic Systems,
vol. 70, no. 1-4, pp. 329–345, 2013.
[3] W. Chang, G. Yang, J. Yu, Z. Liang, L. Cheng, and C. Zhou,
“Development of a power line inspection robot with hybrid operation
modes,” in IEEE/RSJ International Conference on Intelligent Robots
and Systems (IROS). IEEE, 2017, pp. 973–978.
[4] L. Tianyu, L. Yongzhe, Q. Juntong, M. Xiangdong, and H. Jianda,
“Modeling and controller design of hydraulic rotorcraft aerial manip-
ulator,” in 27th Chinese Control and Decision Conference (CCDC).
IEEE, 2015, pp. 5446–5452.
[5] N. Michael, J. Fink, and V. Kumar, “Cooperative manipulation and
transportation with aerial robots,” Autonomous Robots, vol. 30, no. 1,
pp. 73–86, 2011.
[6] J. Cacace, A. Finzi, V. Lippiello, G. Loianno, and D. Sanzone, “Aerial
service vehicles for industrial inspection: task decomposition and
plan execution,Applied Intelligence, vol. 42, no. 1, pp. 49–62, 2015.
[7] S. Shimahara, S. Leewiwatwong, R. Ladig, and K. Shimonomura,
“Aerial torsional manipulation employing multi-rotor flying robot,
in 2016 IEEE/RSJ International Conference on Intelligent Robots
and Systems (IROS). IEEE, 2016, pp. 1595–1600.
[8] P. S´
anchez-Cuevas, G. Heredia, and A. Ollero, “Multirotor UAS for
bridge inspection by contact using the ceiling effect,” in International
Conference on Unmanned Aircraft Systems (ICUAS). IEEE, 2017,
pp. 767–774.
[9] J. R. Kutia, K. A. Stol, and W. Xu, “Aerial manipulator interac-
tions with trees for canopy sampling,” IEEE/ASME Transactions on
Mechatronics, vol. 23, no. 4, pp. 1740–1749, 2018.
[10] J. M. G´
omez-de Gabriel, J. M. Gandarias, F. J. P´
F. J. Garc´
ncz, E. J. Fern´
ıa, and A. J. Garc´
“Methods for autonomous wristband placement with a search-and-
rescue aerial manipulator,” in 2018 IEEE/RSJ International Confer-
ence on Intelligent Robots and Systems (IROS). IEEE, 2018, pp.
[11] N. Staub, M. Mohammadi, D. Bicego, D. Prattichizzo, and
A. Franchi, “Towards robotic magmas: Multiple aerial-ground ma-
nipulator systems,” in IEEE International Conference on Robotics
and Automation (ICRA). IEEE, 2017, pp. 1307–1312.
[12] J. K. Stolaroff, C. Samaras, E. R. ONeill, A. Lubers, A. S. Mitchell,
and D. Ceperley, “Energy use and life cycle greenhouse gas emissions
of drones for commercial package delivery,” Nature communications,
vol. 9, no. 1, p. 409, 2018.
[13] J. Est´
evez, J. M. Lopez-Guede, and M. Gra˜
na, “Quasi-stationary state
transportation of a hose with quadrotors,” Robotics and Autonomous
Systems, vol. 63, pp. 187–194, 2015.
[14] F. Augugliaro, S. Lupashin, M. Hamer, C. Male, M. Hehn, M. W.
Mueller, J. S. Willmann, F. Gramazio, M. Kohler, and R. D’Andrea,
“The flight assembled architecture installation: Cooperative construc-
tion with flying machines,” IEEE Control Systems, vol. 34, no. 4, pp.
46–64, 2014.
[15] C. D. Bellicoso, L. R. Buonocore, V. Lippiello, and B. Siciliano,
“Design, modeling and control of a 5-DoF light-weight robot arm for
aerial manipulation,” in 23rd Mediterranean Conference on Control
and Automation (MED). IEEE, 2015, pp. 853–858.
[16] K. Kondak, F. Huber, M. Schwarzbach, M. Laiacker, D. Sommer,
M. B´
ejar, and A. Ollero, “Aerial manipulation robot composed of
an autonomous helicopter and a 7 degrees of freedom industrial
manipulator,” in 2014 IEEE International Conference on Robotics
and Automation (ICRA). IEEE, 2014, pp. 2107–2112.
[17] C. M. Korpela, T. W. Danko, and P. Y. Oh, “MM-UAV: Mobile
manipulating unmanned aerial vehicle,” Journal of Intelligent and
Robotic Systems: Theory and Applications, vol. 65, no. 1-4, pp. 93–
101, 2012.
[18] C. Korpela, M. Orsag, and P. Oh, “Towards valve turning using
a dual-arm aerial manipulator,” in 2014 IEEE/RSJ International
Conference on Intelligent Robots and Systems (IROS 2014). IEEE,
2014, pp. 3411–3416.
[19] L. M. Matteo Fumagalli, Roberto Naldi, Alessandro Macchelli,
Francesco Forte, Arvid Q.L. Keemink, Stefano Stramigioli, Raffaella
Carloni, “Developing an Aerial Manupulator Prototype,Robotics
and Automation, vol. 21, no. 3, pp. 41–50, 2014.
[20] T. W. Danko and P. Y. Oh, “Toward coordinated manipulator-
host visual servoing for mobile manipulating UAVs,” in Interna-
tional Conference on Technologies for Practical Robot Applications
(TePRA). IEEE, 2014, pp. 1–6.
[21] F. Ruggiero, M. A. Trujillo, R. Cano, H. Ascorbe, A. Viguria,
C. Per´
ez, V. Lippiello, A. Ollero, and B. Siciliano, “A multilayer
control for multirotor UAVs equipped with a servo robot arm,” in
2015 IEEE international conference on robotics and automation
(ICRA). IEEE, 2015, pp. 4014–4020.
[22] B. Theys, G. Dimitriadis, P. Hendrick, and J. De Schutter, “Influence
of propeller configuration on propulsion system efficiency of multi-
rotor Unmanned Aerial Vehicles,” 2016 International Conference on
Unmanned Aircraft Systems, ICUAS 2016, no. June, pp. 195–201,
[23] W. H. Meier and J. R. OLSON, “Efficient sizing of a cargo rotor-
craft,” Journal of Aircraft, vol. 25, no. 6, pp. 538–543, 1988.
[24] M. Gassner, T. Cieslewski, and D. Scaramuzza, “Dynamic collab-
oration without communication: Vision-based cable-suspended load
transport with two quadrotors,” in IEEE International Conference on
Robotics and Automation (ICRA). IEEE, 2017, pp. 5196–5202.
[25] D. Mellinger and V. Kumar, “Minimum snap trajectory generation
and control for quadrotors,” in IEEE International Conference on
Robotics and Automation (ICRA). IEEE, 2011, pp. 2520–2525.
[26] F. Morbidi, R. Cano, and D. Lara, “Minimum-energy path generation
for a quadrotor uav,” in 2016 IEEE International Conference on
Robotics and Automation (ICRA). IEEE, 2016, pp. 1492–1498.
[27] S. Park, J. Her, J. Kim, and D. Lee, “Design, modeling and control
of omni-directional aerial robot,” in International Conference on
Intelligent Robots and Systems (IROS). IEEE, 2016, pp. 1570–1575.
[28] M. Ryll, G. Muscio, F. Pierri, E. Cataldi, G. Antonelli, F. Caccavale,
and A. Franchi, “6D physical interaction with a fully actuated
aerial robot,” in IEEE International Conference on Robotics and
Automation (ICRA). IEEE, 2017, pp. 5190–5195.
[29] T. Kuci´
nski, T. Rybus, K. Seweryn, M. Banaszkiewicz, T. Bura-
towski, G. Chmaj, J. Grygorczuk, and T. Uhl, “Deployable manip-
ulator technology with application for uavs,” in Aerospace Robotics
II. Springer, 2015, pp. 93–103.
[30] B. Yang, Y. He, J. Han, and G. Liu, “Rotor-flying manipulator: mod-
eling, analysis, and control,” Mathematical Problems in Engineering,
vol. 2014, 2014.
[31] I. Maza, K. Kondak, M. Bernard, and A. Ollero, “Multi-UAV
cooperation and control for load transportation and deployment,”
Journal of Intelligent and Robotic Systems: Theory and Applications,
vol. 57, no. 1-4, pp. 417–449, 2010.
[32] P. E. I. Pounds, D. R. Bersak, and A. M. Dollar, “Stability of small-
scale UAV helicopters and quadrotors with added payload mass under
PID control,” Autonomous Robots, vol. 33, no. 1-2, pp. 129–142,
[33] K. Kondak, K. Krieger, A. Albu-Schaeffer, M. Schwarzbach, M. La-
iacker, I. Maza, A. Rodriguez-Castano, and A. Ollero, “Closed-loop
behavior of an autonomous helicopter equipped with a robotic arm
for aerial manipulation tasks,” International Journal of Advanced
Robotic Systems, vol. 10, 2013.
[34] K. Kondak, A. Ollero, I. Maza, K. Krieger, A. Albu-Schaeffer,
M. Schwarzbach, and M. Laiacker, “Unmanned aerial systems
physically interacting with the environment: Load transportation,
deployment, and aerial manipulation,” in Handbook of Unmanned
Aerial Vehicles. Springer, 2015, pp. 2755–2785.
[35] M. Laiacker, F. Huber, and K. Kondak, “High accuracy visual servo-
ing for aerial manipulation using a 7 degrees of freedom industrial
manipulator,IEEE International Conference on Intelligent Robots
and Systems, vol. 2016-November, pp. 1631–1636, 2016.
[36] F. Huber, K. Kondak, K. Krieger, D. Sommer, M. Schwarzbach,
M. Laiacker, I. Kossyk, S. Parusel, S. Haddadin, and A. Albu-
affer, “First analysis and experiments in aerial manipulation using
fully actuated redundant robot arm,” in 2013 IEEE/RSJ International
Conference on Intelligent Robots and Systems (IROS). IEEE, 2013,
pp. 3452–3457.
[37] V. Nayak, C. Papachristos, and K. Alexis, “Design and control of
an aerial manipulator for contact-based inspection,” arXiv preprint
arXiv:1804.03756, 2018.
[38] A. Jimenez-Cano, J. Braga, G. Heredia, and A. Ollero, “Aerial
manipulator for structure inspection by contact from the underside,”
in IEEE/RSJ International Conference on Intelligent Robots and
Systems (IROS). IEEE, 2015, pp. 1879–1884.
[39] K. Karydis and V. Kumar, “Energetics in robotic flight at small
scales,” Interface focus, vol. 7, no. 1, p. 20160088, 2017.
[40] C. Korpela, M. Orsag, T. Danko, B. Kobe, C. McNeil, R. Pisch, and
P. Oh, “Flight stability in aerial redundant manipulators,” in 2012
IEEE International Conference on Robotics and Automation (ICRA).
IEEE, 2012, pp. 3529–3530.
[41] H.-N. Nguyen, S. Park, J. Park, and D. Lee, “A novel robotic
platform for aerial manipulation using quadrotors as rotating thrust
generators,” IEEE Transactions on Robotics, vol. 34, no. 2, pp. 353–
369, 2018.
[42] S. Morton and N. Papanikolopoulos, “A small hybrid ground-air ve-
hicle concept,” in IEEE/RSJ International Conference on Intelligent
Robots and Systems (IROS). IEEE, 2017, pp. 5149–5154.
[43] K. Zhang, P. Chermprayong, T. Alhinai, R. Siddall, and M. Kovac,
“Spidermav: Perching and stabilizing micro aerial vehicles with
bio-inspired tensile anchoring systems,” in IEEE/RSJ International
Conference on Intelligent Robots and Systems (IROS). IEEE, 2017,
pp. 6849–6854.
[44] M. Zhao, K. Kawasaki, X. Chen, S. Noda, K. Okada, and M. Inaba,
“Whole-body aerial manipulation by transformable multirotor with
two-dimensional multilinks,” in IEEE International Conference on
Robotics and Automation (ICRA). IEEE, 2017, pp. 5175–5182.
[45] R. Oung and R. DAndrea, “The distributed flight array: Design,
implementation, and analysis of a modular vertical take-off and
landing vehicle,” The International Journal of Robotics Research,
vol. 33, no. 3, pp. 375–400, 2014.
[46] A. Nikou, G. C. Gavridis, and K. J. Kyriakopoulos, “Mechanical
design, modelling and control of a novel aerial manipulator,” in 2015
IEEE International Conference on Robotics and Automation (ICRA).
IEEE, 2015, pp. 4698–4703.
[47] Smq platform: A novel robotic platform for aerial manipulation.
[Online]. Available:
[48] J. Economou, G. Kladis, A. Tsourdos, and B. White, “UAV optimum
energy assignment using dijkstra’s algorithm,” in European Control
Conference (ECC). IEEE, 2007, pp. 287–292.
[49] M. M. Maia, D. A. Mercado, and F. J. Diez, “Design and imple-
mentation of multirotor aerial-underwater vehicles with experimental
results,” in IEEE/RSJ International Conference on Intelligent Robots
and Systems (IROS). IEEE, 2017, pp. 961–966.
[50] C. Korpela, M. Orsag, Y. Jun, P. Brahmbhatt, and P. Oh, “A hardware-
in-the-loop test rig for aerial manipulation,” in 2013 International
Conference on Unmanned Aircraft Systems (ICUAS). IEEE, 2013,
pp. 982–987.
[51] T. W. Danko and P. Y. Oh, “Evaluation of visual servoing control
of aerial manipulators using test gantry emulation,” in 2014 Interna-
tional Conference on Unmanned Aircraft Systems (ICUAS). IEEE,
2014, pp. 821–829.
[52] D. Kim and P. Y. Oh, “Lab automation drones for mobile manip-
ulation in high throughput systems,” in 2018 IEEE International
Conference on Consumer Electronics (ICCE). IEEE, 2018, pp. 1–5.
[53] G. B. Haberfeld, D. Sun, and N. Hovakimyan, “Stabilization and op-
timal trajectory generation for a compact aerial manipulation system
with a delta-type parallel robot,” in 2018 International Conference on
Unmanned Aircraft Systems (ICUAS). IEEE, 2018, pp. 1091–1100.
[54] P. E. Pounds and A. M. Dollar, “Aerial grasping from a helicopter
UAV platform,” in Experimental Robotics. Springer, 2014, pp. 269–
[55] A. Suarez, G. Heredia, and A. Ollero, “Compliant and lightweight
anthropomorphic finger module for aerial manipulation and grasp-
ing,” in Robot 2015: Second Iberian Robotics Conference. Springer,
2016, pp. 543–555.
[56] A. Suarez, G. Heredia, and R. Ollero, “Lightweight compliant arm
with compliant finger for aerial manipulation and inspection,” in
2016 IEEE/RSJ International Conference on Intelligent Robots and
Systems (IROS). IEEE, 2016, pp. 4449–4454.
[57] S. B. Backus, L. U. Odhner, and A. M. Dollar, “Design of hands for
aerial manipulation: Actuator number and routing for grasping and
perching,” in 2014 IEEE/RSJ International Conference on Intelligent
Robots and Systems. IEEE, 2014, pp. 34–40.
[58] S. B. Backus and A. M. Dollar, “A prismatic-revolute-revolute joint
hand for grasping from unmanned aerial vehicles and other minimally
constrained vehicles,” Journal of Mechanisms and Robotics, vol. 10,
no. 2, p. 025006, 2018.
[59] L. Kruse and J. Bradley, “A hybrid, actively compliant manipula-
tor/gripper for aerial manipulation with a multicopter,” in 2018 IEEE
International Symposium on Safety, Security, and Rescue Robotics
(SSRR). IEEE, 2018, pp. 1–8.
[60] D. Mellinger, Q. Lindsey, M. Shomin, and V. Kumar, “Design, mod-
eling, estimation and control for aerial grasping and manipulation,” in
2011 IEEE/RSJ International Conference on Intelligent Robots and
Systems (IROS). IEEE, 2011, pp. 2668–2673.
[61] D. Mellinger, M. Shomin, N. Michael, and V. Kumar, “Cooperative
grasping and transport using multiple quadrotors,” in Distributed
autonomous robotic systems. Springer, 2013, pp. 545–558.
[62] M. Orsag, C. Korpela, M. Pekala, and P. Oh, “Stability control in
aerial manipulation,” in American Control Conference (ACC). IEEE,
2013, pp. 5581–5586.
[63] A. Gawel, M. Kamel, T. Novkovic, J. Widauer, D. Schindler, B. P.
von Altishofen, R. Siegwart, and J. Nieto, “Aerial picking and
delivery of magnetic objects with MAVs,” in IEEE International
Conference on Robotics and Automation (ICRA). IEEE, 2017, pp.
[64] S. Kim, S. Choi, H. Lee, and H. J. Kim, “Vision-based collaborative
lifting using quadrotor UAVs,” International Conference on Control,
Automation and Systems, no. 2, pp. 1169–1174, 2014.
[65] Nicadrone. [Online]. Available:
[66] N. Toumi, “Decentralized control of multi-agent aerial transportation
system,” 2017.
[67] U. A. Fiaz, M. Abdelkader, and J. S. Shamma, “An intelligent gripper
design for autonomous aerial transport with passive magnetic grasp-
ing and dual-impulsive release,” in 2018 IEEE/ASME International
Conference on Advanced Intelligent Mechatronics (AIM). IEEE,
2018, pp. 1027–1032.
[68] C. C. Kessens, J. Thomas, J. P. Desai, and V. Kumar, “Versatile aerial
grasping using self-sealing suction,” in 2016 IEEE International
Conference on Robotics and Automation (ICRA). IEEE, 2016, pp.
[69] M. Kovaˇ
c, J. Germann, C. H¨
urzeler, R. Y. Siegwart, and D. Floreano,
“A perching mechanism for micro aerial vehicles,” Journal of Micro-
Nano Mechatronics, vol. 5, no. 3-4, pp. 77–91, 2009.
[70] M. Graule, P. Chirarattananon, S. Fuller, N. Jafferis, K. Ma,
M. Spenko, R. Kornbluh, and R. Wood, “Perching and takeoff of a
robotic insect on overhangs using switchable electrostatic adhesion,
Science, vol. 352, no. 6288, pp. 978–982, 2016.
[71] G. Heredia, A. Jimenez-Cano, I. Sanchez, D. Llorente, V. Vega,
J. Braga, J. Acosta, and A. Ollero, “Control of a multirotor outdoor
aerial manipulator,” in International Conference on Intelligent Robots
and Systems (IROS),. IEEE, 2014, pp. 3417–3422.
[72] T. W. Danko and P. Y. Oh, “Design and control of a hyper-redundant
manipulator for mobile manipulating unmanned aerial vehicles,”
Journal of Intelligent and Robotic Systems: Theory and Applications,
vol. 73, no. 1-4, pp. 709–723, 2014.
[73] T. Bartelds, A. Capra, S. Hamaza, S. Stramigioli, and M. Fumagalli,
“Compliant Aerial Manipulators: Toward a New Generation of Aerial
Robotic Workers,IEEE Robotics and Automation Letters, vol. 1,
no. 1, pp. 477–483, 2016.
[74] A. Suarez, G. Heredia, and A. Ollero, “Lightweight compliant arm
for aerial manipulation,” in IEEE/RSJ International Conference on
Intelligent Robots and Systems (IROS). IEEE, 2015, pp. 1627–1632.
[75] A. Q. Keemink, M. Fumagalli, S. Stramigioli, and R. Carloni,
“Mechanical design of a manipulation system for unmanned aerial
vehicles,” in International Conference on Robotics and Automation
(ICRA). IEEE, 2012, pp. 3147–3152.
[76] J. W. Yeol, D. Toohey, and Y.-W. Hwang, “Design and analysis of
a multiple tentacle system for mobile manipulation in micro aerial
vehicles,” Procedia Computer Science, vol. 105, pp. 7–13, 2017.
[77] B. Y¨
uksel, G. Buondonno, and A. Franchi, “Differential flatness
and control of protocentric aerial manipulators with any number of
arms and mixed rigid-/elastic-joints,” in International Conference on
Intelligent Robots and Systems (IROS). IEEE, 2016, pp. 561–566.
[78] J. Thomas, J. Polin, K. Sreenath, and V. Kumar, “Avian-inspired
grasping for quadrotor micro UAVs,” in ASME 2013 international
design engineering technical conferences and computers and infor-
mation in engineering conference. American Society of Mechanical
Engineers, 2013, pp. V06AT07A014–V06AT07A014.
[79] S. Kim, S. Choi, and H. J. Kim, “Aerial manipulation using a quadro-
tor with a two DoF robotic arm,” in 2013 IEEE/RSJ International
Conference on Intelligent Robots and Systems (IROS). IEEE, 2013,
pp. 4990–4995.
[80] G. Garimella and M. Kobilarov, “Towards model-predictive control
for aerial pick-and-place,” in 2015 IEEE International Conference on
Robotics and Automation (ICRA). IEEE, 2015, pp. 4692–4697.
[81] S. Kim, H. Seo, and H. J. Kim, “Operating an unknown drawer
using an aerial manipulator,” in 2015 IEEE International Conference
on Robotics and Automation (ICRA). IEEE, 2015, pp. 5503–5508.
[82] S. Kim, H. Seo, S. Choi, and H. J. Kim, “Vision-guided aerial
manipulation using a multirotor with a robotic arm,” IEEE/ASME
Transactions on Mechatronics, vol. 21, no. 4, pp. 1912–1923, 2016.
[83] H. Seo, S. Kim, and H. J. Kim, “Aerial grasping of cylindrical object
using visual servoing based on stochastic model predictive control,
in IEEE International Conference on Robotics and Automation
(ICRA). IEEE, 2017, pp. 6362–6368.
[84] A. Suarez, P. R. Soria, G. Heredia, B. C. Arrue, and A. Ollero, “An-
thropomorphic, compliant and lightweight dual arm system for aerial
manipulation,” in IEEE/RSJ International Conference on Intelligent
Robots and Systems (IROS). IEEE, 2017, pp. 992–997.
[85] Y. Ohnishi, T. Takaki, T. Aoyama, and I. Ishii, “Development of a
4-joint 3-DOF robotic arm with anti-reaction force mechanism for
a multicopter,” in IEEE/RSJ International Conference on Intelligent
Robots and Systems (IROS). IEEE, 2017, pp. 985–991.
[86] T. Ikeda, S. Yasui, M. Fujihara, K. Ohara, S. Ashizawa, A. Ichikawa,
A. Okino, T. Oomichi, and T. Fukuda, “Wall contact by octo-
rotor UAV with one DoF manipulator for bridge inspection,” in
International Conference on Intelligent Robots and Systems (IROS).
IEEE, 2017, pp. 5122–5127.
[87] M. Fumagalli, S. Stramigioli, and R. Carloni, “Mechatronic design
of a robotic manipulator for unmanned aerial vehicles,” in 2016
IEEE/RSJ International Conference on Intelligent Robots and Sys-
tems (IROS). IEEE, 2016, pp. 4843–4848.
[88] M. Kamel, K. Alexis, and R. Siegwart, “Design and modeling
of dexterous aerial manipulator,” in 2016 IEEE/RSJ International
Conference on Intelligent Robots and Systems (IROS). IEEE, 2016,
pp. 4870–4876.
[89] J. L. Scholten, M. Fumagalli, S. Stramigioli, and R. Carloni, “In-
teraction control of an uav endowed with a manipulator,” in 2013
IEEE International Conference on Robotics and Automation (ICRA).
IEEE, 2013, pp. 4910–4915.
[90] J. Fink, N. Michael, S. Kim, and V. Kumar, “Planning and control
for cooperative manipulation and transportation with aerial robots,
The International Journal of Robotics Research, vol. 30, no. 3, pp.
324–334, 2011.
[91] Q. Jiang and V. Kumar, “The inverse kinematics of cooperative
transport with multiple aerial robots,” IEEE Transactions on Robotics,
vol. 29, no. 1, pp. 136–145, 2013.
[92] V. Spurn `
y, T. B´
ca, M. Saska, R. Pˇ
cka, T. Krajn´
ık, J. Thomas,
D. Thakur, G. Loianno, and V. Kumar, “Cooperative autonomous
search, grasping, and delivering in a treasure hunt scenario by a team
of unmanned aerial vehicles,” Journal of Field Robotics, vol. 36,
no. 1, pp. 125–148, 2019.
[93] R. Spica, A. Franchi, G. Oriolo, H. H. B¨
ulthoff, and P. R. Giordano,
“Aerial grasping of a moving target with a quadrotor uav,” in
2012 IEEE/RSJ International Conference on Intelligent Robots and
Systems. IEEE, 2012, pp. 4985–4992.
[94] I. Palunko, A. Faust, P. Cruz, L. Tapia, and R. Fierro, “A reinforce-
ment learning approach towards autonomous suspended load manip-
ulation using aerial robots,” in 2013 IEEE International Conference
on Robotics and Automation. IEEE, 2013, pp. 4896–4901.
[95] P. Ramon Soria, R. Bevec, B. C. Arrue, A. Ude, and A. Ollero,
“Extracting objects for aerial manipulation on UAVs using low cost
stereo sensors,” Sensors (Switzerland), vol. 16, no. 5, pp. 1–20, 2016.
[96] C. Luo, L. Yu, and P. Ren, “A vision-aided approach to perching a
bioinspired unmanned aerial vehicle,” IEEE Transactions on Indus-
trial Electronics, vol. 65, no. 5, pp. 3976–3984, 2018.
[97] R. Ladig, S. Leewiwatwong, and K. Shimonomura, “Fpga-based fast
response image analysis for orientational control in aerial manipula-
tion tasks,” Journal of Signal Processing Systems, pp. 1–11, 2018.
[98] M. Orsag, C. M. Korpela, S. Bogdan, and P. Y. Oh, “Hybrid adaptive
control for aerial manipulation,” Journal of intelligent & robotic
systems, vol. 73, no. 1-4, pp. 693–707, 2014.
[99] D. Bazylev, A. Kremlev, A. Margun, and K. Zimenko, “Design of
control system for a four-rotor UAV equipped with robotic arm,
in 7th International Congress on Ultra Modern Telecommunications
and Control Systems and Workshops (ICUMT). IEEE, 2015, pp.
[100] D. Bazylev, A. Margun, K. Zimenko, and A. Kremlev, “UAV
equipped with a robotic manipulator,” in 22nd Mediterranean Con-
ference on Control and Automation. IEEE, 2014, pp. 1177–1182.
[101] J. P. S. Buzzato, A. C. Hernandes, M. Becker, and G. A.
de Paula Caurin, “Aerial manipulation with six-axis force and torque
sensor feedback compensation,” in Latin American Robotic Sympo-
sium, Brazilian Symposium on Robotics (SBR) and Workshop on
Robotics in Education (WRE). IEEE, 2018, pp. 158–163.
[102] M. Orsag, C. Korpela, S. Bogdan, and P. Oh, “Lyapunov based
model reference adaptive control for aerial manipulation,” in 2013
International Conference on Unmanned Aircraft Systems (ICUAS).
IEEE, 2013, pp. 966–973.
[103] S. Di Lucia, G. D. Tipaldi, and W. Burgard, “Attitude stabilization
control of an aerial manipulator using a quaternion-based backstep-
ping approach,” in European Conference on Mobile Robots (ECMR).
IEEE, 2015, pp. 1–6.
[104] J. Acosta, M. Sanchez, and A. Ollero, “Robust control of underactu-
ated aerial manipulators via IDA-PBC,” in 53rd Annual Conference
on Decision and Control (CDC),. IEEE, 2014, pp. 673–678.
[105] A. Khalifa, M. Fanni, A. Ramadan, and A. Abo-Ismail, “Adaptive
intelligent controller design for a new quadrotor manipulation sys-
tem,” in 2013 IEEE International Conference on Systems, Man, and
Cybernetics. IEEE, 2013, pp. 1666–1671.
[106] A. Jimenez-Cano, J. Martin, G. Heredia, A. Ollero, and R. Cano,
“Control of an aerial robot with multi-link arm for assembly tasks,”
in International Conference on Robotics and Automation. IEEE,
2013, pp. 4916–4921.
[107] A. Y. Mersha, S. Stramigioli, and R. Carloni, “Exploiting the dynam-
ics of a robotic manipulator for control of UAVs,” in International
Conference on Robotics and Automation (ICRA). IEEE, 2014, pp.
[108] S. Kannan, M. Alma, M. A. Olivares-Mendez, and H. Voos, “Adap-
tive control of aerial manipulation vehicle,” in International Con-
ference on Control System, Computing and Engineering (ICCSCE
2014). IEEE, 2014, pp. 273–278.
[109] V. Ghadiok, J. Goldin, and W. Ren, “On the design and development
of attitude stabilization, vision-based navigation, and aerial gripping
for a low-cost quadrotor,” Autonomous Robots, vol. 33, no. 1-2, pp.
41–68, 2012.
[110] G. Muscio, F. Pierri, M. A. Trujillo, E. Cataldi, G. Giglio, G. An-
tonelli, F. Caccavale, A. Viguria, S. Chiaverini, and A. Ollero,
“Experiments on coordinated motion of aerial robotic manipulators,”
in 2016 IEEE international conference on robotics and automation
(ICRA). IEEE, 2016, pp. 1224–1229.
[111] F. Caccavale, G. Giglio, G. Muscio, and F. Pierri, “Cooperative
impedance control for multiple uavs with a robotic arm,” in IEEE/RSJ
International Conference on Intelligent Robots and Systems (IROS).
IEEE, 2015, pp. 2366–2371.
[112] E. Cataldi, G. Muscio, M. A. Trujillo, Y. Rodr´
ıguez, F. Pierri,
G. Antonelli, F. Caccavale, A. Viguria, S. Chiaverini, and A. Ollero,
“Impedance control of an aerial-manipulator: Preliminary results,” in
2016 IEEE/RSJ International Conference on Intelligent Robots and
Systems (IROS). IEEE, 2016, pp. 3848–3853.
[113] M. Sharifi and H. Sayyaadi, “Nonlinear robust adaptive Cartesian
impedance control of UAVs equipped with a robot manipulator,
Advanced Robotics, vol. 29, no. 3, pp. 171–186, 2015.
[114] A. Suarez, G. Heredia, and A. Ollero, “Physical-virtual impedance
control in ultralightweight and compliant dual-arm aerial manipu-
lators,” IEEE Robotics and Automation Letters, vol. 3, no. 3, pp.
2553–2560, 2018.
[115] S. Hamaza, I. Georgilas, and T. Richardson, “An adaptive-
compliance manipulator for contact-based aerial applications,” in
2018 IEEE/ASME International Conference on Advanced Intelligent
Mechatronics (AIM). IEEE, 2018, pp. 730–735.
[116] G. Giglio and F. Pierri, “Selective compliance control for an un-
manned aerial vehicle with a robotic arm,” in 22nd Mediterranean
Conference on Control and Automation. IEEE, 2014, pp. 1190–
[117] H. Sayyaadi and M. Sharifi, “Adaptive impedance control of uavs
interacting with environment using a robot manipulator,” in 2014
Second RSI ISM International Conference on Robotics and Mecha-
tronics (ICRoM). IEEE, 2014, pp. 636–641.
[118] M. Srikanth, A. Soto, A. Annaswamy, E. Lavretsky, and J.-J. Slo-
tine, “Controlled manipulation with multiple quadrotors,” in AIAA
Guidance, Navigation, and Control Conference, 2011, pp. 65–47.
[119] R. Mebarki, V. Lippiello, and B. Siciliano, “Image-based control
for dynamically cross-coupled aerial manipulation,” in International
Conference on Intelligent Robots and Systems (IROS). IEEE, 2014,
pp. 4827–4833.
[120] L. R. Buonocore, J. Cacace, and V. Lippiello, “Hybrid visual servoing
for aerial grasping with hierarchical task-priority control,” in 2015
23rd Mediterranean Conference on Control and Automation (MED).
IEEE, 2015, pp. 617–623.
[121] R.˜Mebarki and V.˜Lippiello, “Image-based control for aerial manip-
ulation,” Asian Journal of Control, vol. 16, no. 3, pp. 646–656, 2014.
[122] L. Fang, H. Chen, Y. Lou, Y. Li, and Y. Liu, “Visual grasping
for a lightweight aerial manipulator based on nsga-ii and kinematic
compensation,” in 2018 IEEE International Conference on Robotics
and Automation (ICRA). IEEE, 2018, pp. 1–6.
[123] V. Lippiello, G. A. Fontanelli, and F. Ruggiero, “Image-based visual-
impedance control of a dual-arm aerial manipulator,IEEE Robotics
and Automation Letters, vol. 3, no. 3, pp. 1856–1863, 2018.
[124] K. P. Valavanis and G. J. Vachtsevanos, “Handbook of unmanned
aerial vehicles,” Handbook of Unmanned Aerial Vehicles, pp. 1–3022,
[125] H. Lee, H. Kim, and H. J. Kim, “Planning and control for collision-
free cooperative aerial transportation,IEEE Transactions on Au-
tomation Science and Engineering, 2016.
[126] G. Loianno and V. Kumar, “Cooperative transportation using small
quadrotors using monocular vision and inertial sensing,” IEEE
Robotics and Automation Letters, vol. 3, no. 2, pp. 680–687, 2018.
[127] H. Lee and H. J. Kim, “Constraint-based cooperative control of
multiple aerial manipulators for handling an unknown payload,” IEEE
Transactions on Industrial Informatics, vol. 13, no. 6, pp. 2780–2790,
[128] A. Tagliabue, M. Kamel, S. Verling, R. Siegwart, and J. Nieto, “Col-
laborative transportation using MAVs via passive force control,” in
IEEE International Conference on Robotics and Automation (ICRA).
IEEE, 2017, pp. 5766–5773.
[129] M. Tognon, C. Gabellieri, L. Pallottino, and A. Franchi, “Aerial co-
manipulation with cables: The role of internal force for equilibria,
stability, and passivity,” IEEE Robotics and Automation Letters,
vol. 3, no. 3, pp. 2577–2583, 2018.
[130] C. Masone, H. H. B¨
ulthoff, and P. Stegagno, “Cooperative trans-
portation of a payload using quadrotors: A reconfigurable cable-
driven parallel robot,” in 2016 IEEE/RSJ International Conference
on Intelligent Robots and Systems (IROS). IEEE, 2016, pp. 1623–
[131] J. E. Sanz, “Quadrotor Team Modeling and Control for DLO Trans-
portation,” 2016.
[132] I. H. B. Pizetta, A. S. Brandao, and M. Sarcinelli-Filho, “Cooperative
quadrotors carrying a suspended load,” 2016 International Confer-
ence on Unmanned Aircraft Systems, ICUAS 2016, pp. 1049–1055,
[133] G. Gioioso, A. Franchi, G. Salvietti, S. Scheggi, and D. Prattichizzo,
“The flying hand: A formation of uavs for cooperative aerial tele-
manipulation,” in 2014 IEEE International Conference on Robotics
and Automation (ICRA). IEEE, 2014, pp. 4335–4341.
[134] M. Mohammadi, A. Franchi, D. Barcelli, and D. Prattichizzo, “Co-
operative aerial tele-manipulation with haptic feedback,” in 2016
IEEE/RSJ International Conference on Intelligent Robots and Sys-
tems (IROS). IEEE, 2016, pp. 5092–5098.
[135] D. Montufar, F. Munoz, E. Espinoza, O. Garcia, and S. Salazar,
“Multi-uav testbed for aerial manipulation applications,” in 2014
International Conference on Unmanned Aircraft Systems (ICUAS).
IEEE, 2014, pp. 830–835.
[136] J. Bannwarth, Z. Chen, K. Stol, and B. MacDonald, “Disturbance
accomodation control for wind rejection of a quadcopter,” in 2016
International Conference on Unmanned Aircraft Systems (ICUAS).
IEEE, 2016, pp. 695–701.
[137] Y. Zhang, Z. Chen, X. Zhang, Q. Sun, and M. Sun, “A novel control
scheme for quadrotor uav based upon active disturbance rejection
control,” Aerospace Science and Technology, vol. 79, pp. 601–609,
[138] H. Paul, K. Ono, R. Ladig, and K. Shimonomura, “A multirotor
platform employing a three-axis vertical articulated robotic arm
for aerial manipulation tasks,” in 2018 IEEE/ASME International
Conference on Advanced Intelligent Mechatronics (AIM). IEEE,
2018, pp. 478–485.
[139] C. E. Doyle, J. J. Bird, T. A. Isom, C. J. Johnson, J. C. Kallman, J. A.
Simpson, R. J. King, J. J. Abbott, and M. A. Minor, “Avian-inspired
passive perching mechanism for robotic rotorcraft,” in IEEE/RSJ
International Conference on Intelligent Robots and Systems (IROS).
IEEE, 2011, pp. 4975–4980.
[140] A. Abdilla, A. Richards, and S. Burrow, “Power and endurance mod-
elling of battery-powered rotorcraft,” in 2015 IEEE/RSJ International
Conference on Intelligent Robots and Systems (IROS). IEEE, 2015,
pp. 675–680.
[141] J. Estevez and M. Gra˜
na, “Improved Control of DLO Transportation
by a Team of Quadrotors,” in International Work-Conference on the
Interplay Between Natural and Artificial Computation. Springer,
2017, pp. 117–126.
[142] J. Gimenez, D. C. Gandolfo, L. R. Salinas, C. Rosales, and R. Carelli,
“Multi-objective control for cooperative payload transport with rotor-
craft UAVs,” ISA transactions, 2018.
[143] Y. Wu, J. Song, J. Sun, F. Zhu, and H. Chen, “Aerial Grasping Based
on VR Perception and Haptic Control,” in 2018 IEEE International
Conference on Real-time Computing and Robotics (RCAR). IEEE,
2018, pp. 556–562.
... With the advent of advanced sensor/communication technologies, computer vision, deep learning algorithms, and reliable robotic platforms (Unmanned aerial, ground, surface, underwater vehicles), there has been an increased interest among researchers in the area of cooperative Multi-Robotic Systems (MRS). Since the associated technologies are getting cheaper, smaller, and more reliable, MRS are attractive for usage in unsafe and uncertain environments [1]. MRS have a wide range of applications such as search and rescue [2], firefighting [3], convoy protection [4], traffic monitoring [5], surveillance, etc., and all these applications involve tracking a target as one of the fundamental tasks. ...
... Target Model: The target model is similar to the robot model. The target's position vector x t,Tg ∈ R 2 (in m), heading angle φ t,Tg ∈ R (radians), body-axis velocityv t,Tg ∈ R 2 (m/s), and yaw ratew t,Tg ∈ R (rad/s), respectively, can be represented by replacing i with T g in the set of equations (1). Similarly,v t,Tg andw t,Tg act as bounded control inputs for the target at time t, which are considered unknown to the robots. ...
... The drift term ζ 1 t,i = c 1 t,i s t follows the model given by equations (8) and (9), and the drift reset probability is set to be p = 0.1. Further, we consider the term c 1 t,i = γ t,i [1,1] , and ν 1 t,i 's covariance term C 1 t,i = (10 · γ t,i ) 2 diag( [1,1]), where the terms γ t,i , ∀i ∈ [N ], vary with time as shown in Table I. γ t,i values are indicative of how good or bad algorithm A i is at time t; larger γ t,i values lead to a lower prediction accuracy. From table I, note the variation in the γ t,i over the horizon of T = 1400 discrete-time steps; for instance, algorithm A 1 (installed in the 1 st robot) is accurate initially but its prediction degrades later on, whereas the opposite can be said about algorithm A 6 (installed in the 6 th robot). ...
Full-text available
This paper addresses the problem of cooperative target tracking using a heterogeneous multi-robot system, where the robots are communicating over a dynamic communication network, and heterogeneity is in terms of different types of sensors and prediction algorithms installed in the robots. The problem is cast into a distributed learning framework, where robots are considered as 'agents' connected over a dynamic communication network. Their prediction algorithms are considered as 'experts' giving their look-ahead predictions of the target's trajectory. In this paper, a novel Decentralized Distributed Expert-Assisted Learning (D2EAL) algorithm is proposed, which improves the overall tracking performance by enabling each robot to improve its look-ahead prediction of the target's trajectory by its information sharing, and running a weighted information fusion process combined with online learning of weights based on a prediction loss metric. Theoretical analysis of D2EAL is carried out, which involves the analysis of worst-case bounds on cumulative prediction loss, and weights convergence analysis. Simulation studies show that in adverse scenarios involving large dynamic bias or drift in the expert predictions, D2EAL outperforms well-known covariance-based estimate/prediction fusion methods, both in terms of prediction performance and scalability.
... Unfortunately, most existing UAVs can only perform some perceptional tasks passively while being unable to interact with the environment actively. Hence, researchers have tried to develop aerial manipulators [4][5][6][7], which is composed of UAVs and multiple degrees of freedom manipulators or grippers. Thanks to the dexterous motion capacity of the aircrafts, aerial manipulators can achieve aerial-ground interaction activities such as remote scientific sampling [8], aerial grasping [9], and ground target capture [10]. ...
... Finally, the linear extended state observers for system (6) can be suggested as ˙ z = A z + Bu + L(y −ŷ) y = C z (9) whereŷ is the system output, z = q 1 , q 1 ,x 3 T is the estimation of z,q 1 , q 1 ,x 3 are the estimation of ...
... Due to its concealment of flight and diversity of functions, it can realize both aerial reconnaissance and underwater inspection, which expands the spatial scope of navigation [2]. Therefore, it combines the advantages of aerial drones and underwater submersibles, which have been favored by researchers from various countries since the early 20th century [3]. However, due to the large difference in the characteristics of the water-air medium, it is not a simple matter to cross the water-air medium, which involves the complex model entering and exiting the water process [4]. ...
Full-text available
A trans-medium aircraft is a new concept aircraft that can both dive in the water and fly in the air. In this paper, a new type of water–air multi-medium span vehicle is designed based on the water entry and exit structure model of a multi-rotor UAV. Based on the designed structural model of the cross-media aircraft, the OpenFOAM open source numerical platform is used to analyze the single-medium aerodynamic characteristics and the multi-medium spanning flow analysis. The rotating flow characteristics of single-medium air rotor and underwater propeller are calculated by sliding mesh. In order to prevent the numerical divergence caused by the deformation of the grid movement, the overset grid method and the multiphase flow technology are used for the numerical simulation of the water entry and exit of the cross-medium aircraft. Through the above analysis, the flow field characteristics of the trans-medium vehicle in different media are verified, and the changes in the body load and attitude at different water entry angles are also obtained during the process of medium crossing.
... The progress of aerial manipulation has dramatically accelerated in the past few years [23,42,118,124,160,179]. However, the earliest attempts to apply force to the environment using multirotors go back to 2010. ...
The physical interaction of aerial robots with their environment has countless potential applications and is an emerging area with many open challenges. Fully-actuated multirotors have been introduced to tackle some of these challenges. They provide complete control over position and orientation and eliminate the need for attaching a multi-DoF manipulation arm to the robot. However, there are many open problems before they can be used in real-world applications. Researchers have introduced some methods for physical interaction in limited settings. Their experiments primarily use prototype-level software without an efficient path to integration with real-world applications. We describe a new cost-effective solution for integrating these robots with the existing software and hardware flight systems for real-world applications and expand it to physical interaction applications. On the other hand, the existing control approaches for fully-actuated robots assume conservative limits for the thrusts and moments available to the robot. Using conservative assumptions for these already-inefficient robots makes their interactions even less optimal and may even result in many feasible physical interaction applications becoming infeasible. This work proposes a real-time method for estimating the complete set of instantaneously available forces and moments that robots can use to optimize their physical interaction performance. Finally, many real-world applications where aerial robots can improve the existing manual solutions deal with deformable objects. However, the perception and planning for their manipulation is still challenging. This research explores how aerial physical interaction can be extended to deformable objects. It provides a detection method suitable for manipulating deformable one-dimensional objects and introduces a new perspective on planning the manipulation of these objects.
... UAVs, more commonly addressed as drones, are small-sized aircraft without a pilot that can remotely be controlled or set to fly through software-controlled flight plans in their embedded systems that work in coordination with the sensors associated with the drone and the Global Positioning System (GPS). Some newer technologies like vision-based navigation [11] and ariel flight path manipulation [12] have broadened the array of the application of UAVs. "Approximately 3.55 million UAVs are expected to be deployed for consumer use by 2020 in the United States" [13] and were initially introduced for serving the military in the 20th century for missions considered too "dull, dirty or dangerous" [14] for humans. ...
... A quaternion q with q = 1, element of S 3 , is called a unit quaternion, parameterizing a 4-dimensional unit-sphere, and, contrary to a general quaternion in H, solemnly describe the orientation of a rigid body. S 3 is a double cover of SO (3) meaning that q and −q characterize the same orientation. ...
Full-text available
Aerial manipulators (AM) exhibit particularly challenging, non-linear dynamics; the UAV and the manipulator it is carrying form a tightly coupled dynamic system, mutually impacting each other. The mathematical model describing these dynamics forms the core of many solutions in non-linear control and deep reinforcement learning. Traditionally, the formulation of the dynamics involves Euler angle parametrization in the Lagrangian framework or quaternion parametrization in the Newton-Euler framework. The former has the disadvantage of giving birth to singularities and the latter being algorithmically complex. This work presents a hybrid solution, combining the benefits of both, namely a quaternion approach leveraging the Lagrangian framework, connecting the singularity-free parameterization with the algorithmic simplicity of the Lagrangian approach. We do so by offering detailed insights into the kinematic modeling process and the formulation of the dynamics of a general aerial manipulator. The obtained dynamics model is validated experimentally against a real-time physics engine. A practical application of the obtained dynamics model is shown in the context of a computed torque feedback controller (feedback linearization), where we analyze its real-time capability with increasingly complex models.
In this paper, we conceptualize, analyze, and assemble a prototype adaptive surface system capable of morphing its geometric configuration using an array of linear actuators to impose omnidirectional movement of objects that lie on the surface. The principal focus and contribution of this paper is the derivation of feedback control protocols–for regulating the actuators’ length in order to accomplish the object conveyance task–that scale with the number of actuators and the nonlinear kinematic constraints of the morphing surface. Simulations and experimental results demonstrate the advantages of distributed manipulation over static-shaped feeders.
Full-text available
In recent years, the agricultural sector has turned to robotic automation to deal with the growing demand for food. Harvesting fruits and vegetables is the most labor-intensive and time-consuming among the main agricultural tasks. However, seasonal labor shortage of experienced workers results in low efficiency of harvesting, food losses, and quality deterioration. Therefore, research efforts focus on the automation of manual harvesting operations. Robotic manipulation of delicate products in unstructured environments is challenging. The development of suitable end effectors that meet manipulation requirements is necessary. To that end, this work reviews the state-of-the-art robotic end effectors for harvesting applications. Detachment methods, types of end effectors, and additional sensors are discussed. Performance measures are included to evaluate technologies and determine optimal end effectors for specific crops. Challenges and potential future trends of end effectors in agricultural robotic systems are reported. Research has shown that contact-grasping grippers for fruit holding are the most common type of end effectors. Furthermore, most research is concerned with tomato, apple, and sweet pepper harvesting applications. This work can be used as a guide for up-to-date technology for the selection of suitable end effectors for harvesting robots.
Tony, Lima AgnelJana, ShuvrangshuVarun, V. P.Shorewala, ShantamVidyadhara, B. V.Gadde, Mohitvishnu S.Kashyap, AbhishekRavichandran, RahulKrishnapuram, RaghuGhose, DebasishCapturing moving objects using unmanned aerial vehicles (UAVs) is a challenging task. Many UAV applications require the capture of dynamic aerial targets. Successful interception requires accurate detection of the object, continuous tracking, and safe engagement without damaging any involving vehicles. This work presents the algorithmic details and hardware implementation for capturing a moving ball in a collaborative framework in an outdoor environment. The tracking and grabbing algorithm is developed using image-based guidance from the information of a monocular camera. The target image is detected and tracked using ML-based algorithm and Kalman filter. Finally, the proposed framework is simulated in ROS-Gazebo to evaluate the performance of individual algorithms and further implemented on hardware to validate the system’s real-time performance. The proposed system could be utilized for several applications like counter-UAV systems, fruit picking, among many others.
Conference Paper
Full-text available
A new robotic system for Search And Rescue (SAR) operations based on the automatic wristband placement on the victims’ arm, which may provide identification, beaconing and remote sensor readings for continuous health monitoring. This paper focuses on the development of the automatic target localization and the device placement using an unmanned aerial manipulator. The automatic wrist detection and localization system uses an RGB-D camera and a convolutional neural network based on the region faster method (Faster R-CNN). A lightweight parallel delta manipulator with a large workspace has been built, and a new design of a wristband in the form of a passive detachable gripper, is presented, which under contact, automatically attaches to the human, while disengages from the manipulator. A new trajectory planning method has been used to minimize the torques caused by the external forces during contact, which cause attitude perturbations. Experiments have been done to evaluate the machine learning method for detection and location, and for the assessment of the performance of the trajectory planning method. The results show how the VGG-16 neural network provides a detection accuracy of 67.99%. Moreover, simulation experiments have been done to show that the new trajectories minimize the perturbations to the aerial platform.
Full-text available
This paper addresses the problem of autonomous cooperative localization, grasping and delivering of colored ferrous objects by a team of unmanned aerial vehicles (UAVs). In the proposed scenario, a team of UAVs is required to maximize the reward by collecting colored objects and delivering them to a predefined location. This task consists of several subtasks such as cooperative coverage path planning, object detection and state estimation, UAV self‐localization, precise motion control, trajectory tracking, aerial grasping and dropping, and decentralized team coordination. The failure recovery and synchronization job manager is used to integrate all the presented subtasks together and also to decrease the vulnerability to individual subtask failures in real‐world conditions. The whole system was developed for the Mohamed Bin Zayed International Robotics Challenge (MBZIRC) 2017, where it achieved the highest score and won Challenge No. 3—Treasure Hunt. This paper does not only contain results from the MBZIRC 2017 competition but it also evaluates the system performance in simulations and field tests that were conducted throughout the year‐long development and preparations for the competition.
In this paper, we consider the problem of controlling multiple quadrotor robots that cooperatively grasp and transport a payload in three dimensions.We model the quadrotors both individually and as a group rigidly attached to a payload. We propose individual robot control laws defined with respect to the payload that stabilize the payload along three-dimensional trajectories. We detail the design of a gripping mechanism attached to each quadrotor that permits autonomous grasping of the payload. An experimental study with teams of quadrotors cooperatively grasping, stabilizing, and transporting payloads along desired three-dimensional trajectories is presented with performance analysis over many trials for different payload configurations.