Application of caging manipulation and compliant mechanism for a container case hand-over task
ABSTRACT This research aims to realize a container case hand-over task between robots. Although sophisticated cooperative control is essential to avoid destructive internal force in general cooperative transfer task, we propose a geometrical caging strategy which can simplify a hand-over task. To solve a problem of capture region mismatch during caging state transition, the proposed strategy does not utilize additional actuators and sensors but a compliant mechanism whose displacement range can be changed automatically by its configuration. By experiments, it was confirmed that the caging strategy can make the hand-over task drastically easy and robust, and a winch type quasi-compliant mechanism is effective to the problem of capture region mismatch.
Conference Proceeding: Distributed robot helpers handling a single object in cooperation with a human[show abstract] [hide abstract]
ABSTRACT: In this paper, we propose distributed robot helpers (DR helpers) and a decentralized control algorithm for them to transport a single object in cooperation with a human/humans. The DR helper consists of an omni-directional mobile base, a six-axis body force sensor, a folk lift, and an onboard controller. Each robot is controlled as if it has a caster-like mechanism. The adaptive dual caster action is proposed to improve the maneuverability of the system. Multiple DR helpers could transport a single object in cooperation with a human based on the operator's intentional force/moment. Experiments using multiple DR helpers will illustrate the validity of the proposed control systemRobotics and Automation, 2000. Proceedings. ICRA '00. IEEE International Conference on; 02/2000
Article: Decentralized adaptive coordinated control of multiple robot arms without using a force sensor[show abstract] [hide abstract]
ABSTRACT: This paper presents a distributed adaptive coordinated control method for multiple robot arms grasping a common object. The cases of rigid contact and rolling contact are analyzed. In the proposed controller, the dynamic parameters of both object and robot arms are estimated adaptively. The desired motions of the robot arms are generated by an estimated object reference model. The control method requires only the measurements of the positions and velocities of the object and robot arms, but not the measurements of forces and moments at contact points. The asymptotic convergence of trajectory is proven by the Lyapunov-like Lemma. Experiments involving two robot arms handling a common object are shown.Automatica. 01/2006;
Conference Proceeding: Two-finger caging of concave polygon[show abstract] [hide abstract]
ABSTRACT: An object is captured by a set of fingers when there exists no trajectory to bring the object arbitrarily far from the fingers. Object concavity is a special geometric property that allows objects to be captured with only few fingers. In particular, certain concave objects may be captured by appropriately placing two fingers close to some pair of opposite concave sections. This paper addresses the problem of computing all configurations of the fingers that are farthest away from each other while still capable of capturing the object. We propose an O(n<sup>2</sup>lg n) algorithm for this task and present preliminary results showing efficiency of the algorithmRobotics and Automation, 2006. ICRA 2006. Proceedings 2006 IEEE International Conference on; 06/2006
Application of Caging Manipulation and Compliant Mechanism
for a Container Case Hand-over Task
Rui FUKUI, Taketoshi MORI, Tomomasa SATO.
This research aims to realize a container case hand-
over task between robots. Although sophisticated cooperative
control is essential to avoid destructive internal force in
general cooperative transfer task, we propose a geometrical
caging strategy which can simplify a hand-overtask. To solve
a problem of capture region mismatch during caging state
transition, the proposed strategy does not utilize additional
actuators and sensors but a compliant mechanism whose
displacement range can be changed automatically by its
configuration. By experiments, it was confirmed that the
caging strategy can make the hand-over task drastically easy
and robust, and a winch type quasi-compliant mechanism is
effective to the problem of capture region mismatch.
An object hand-over task between multiple robots is more
complex and difficult than a handling task by single robot.
This is because a sophisticated cooperative control is re-
quired to hand-over an object between robots, and if there is
any failure the object will be deformed or broken. There were
some previous works about cooperative object transferring
task between human and a robot,  or multiple robots,
 . Hirata et al. proposed an idea: “Virtual 3D caster” and
actualized leader-follower type smooth cooperative transfer
motion. In recent researches, multiple robots cooperative
motion without a force sensor were reported, . These
researches presented some solutions to the problem of co-
operative object transfer task which is essential to avoid
destructive internal force.
On the other hand, an object hand-over task has a different
problem from the cooperative transfer task in terms of a
stable state transition. That is to say, the cooperative transfer
task has only one state, but hand-over task must realize
a smooth and stable state transition from a robot grasping
state to another robot grasping state. The transition problem
cannot be solved by the previous approaches.
Therefore this paper discusses a strategy for a stable
object hand-over task which utilizes caging manipulation and
compliant mechanisms. An application target of the proposed
strategy is our developing home-use logistical support robot
system, and the hand-over target is intelligent container (i-
Container): the core element of the system (Fig.1, ). The
Class SClass AClass E
Fig. 1. i-Container family
goal is to achieve i-Container hand-over task between a con-
tainer transfer robot and home-use automated container
The framework of this paper is as follows. In Section
II, the problem of container hand-over task will be defined
and caging manipulation strategy will be discussed. Section
III explains the hand-over task installation environment;
hand-over target i-Container, a container transfer robot and
a home-use automated container storage/retrieval system.
Section IV will show the overall task flow and select optimal
methods at each process. Section V describes performance
experiments of the implemented hand-over task. Section VI
II. CONTAINER HAND-OVER TASK
This section will define difficulty and a problem of an
object hand-over task, and propose a caging manipulation
strategy to overcome the problem. Next, previous works
about caging manipulation will introduced and an optimal
caging configuration for the hand-overtask will be discussed.
A. Problem definition in an object hand-over task
Firstly we suppose an object hand-over task between
Robot A and Robot B, the task is composed of following 3
processes, (1) Stable object manipulation by Robot A alone,
(2) Stable cooperative object handling by both Robot A and
Robot B, (3) Stable object manipulation by Robot B alone.
*Here throwing and catching motion is out of our target,
because the target object can be frequently un-stable.
As described in section I, cooperative handling like
Fig.2(A) was investigated by some previous works. But
transition from cooperative handling to stand alone manipu-
lation like Fig.2(B) cannot be achieved in the framework of
previous works and it may be difficult to expand their theory
to our target task. Therefore we must consider a strategy that
can simplify the hand-over task, and the strategy may be
quite different from the previous approaches.
(A) Stable cooperative
(B) Unstable transition
of hand-over motion
Fig. 2. Difficulty of state transition
(A) Relaying strategy
(B) Caging manipulation strategy
Fig. 3.Conceptual diagram of simplification strategies (Top view)
B. Strategy to simplify the hand-over task
There are two directions to simplify the hand-over task.
1) Relaying strategy (Fig.3(A))
2) Caging manipulation strategy (Fig.3(B))
In the relaying strategy robots execute hand-over motion via
a relaying place, and robots perform each pick and place
motion. Extra space and accurate object position/posture
recognition technologies are essential to apply this strategy,
therefore it’s not feasible in all cases.
Hence caging manipulation strategy can be an appropriate
candidate, in the strategy hand-over target object is con-
strained flexibly by caging. As flexible constraint conditions,
there are several other methods; (1) Frictional support like
fork lift (Fig.4), (2) Accurate positioning free grasping mech-
anism like electro-magnets. But when an accidental contact
or collision occurs, an object in the frictional support can
change its position and posture and a receiver robot needs
to confirm the object state with sophisticated sensor, and
positioning free grasping mechanism generally cannot realize
robust and stable support. Accordingly we decide to utilize
simple and robust geometric constraint.
In the hand-over task, the target object can be applied two
different caging conditions. That means the object must have
two different structures for caging, one is for a transmitter
robot and the other is for a receiver robot.
In general caging strategy can avoid unnecessary internal
force. However flexible constraint condition intends difficulty
of accurate positioning. So a compliant mechanism becomes
a good solution to absorb the position uncertainty in the
caging hand-over task.
Fig. 4.Freight transporting by forklift
C. Related research of caging
In a set-theoretical definition, caging condition can be
describes as equation (1).
free object motion, qobjis a current object configuration, and
Cfree.inf is free object motion configuration space which is
infinitely far away from the current configuration.
That means 1st equation ensures existence of caging state
itself, and the 2nd equation ensures existence of capture
region. (i.e. this eliminates a situation where an object is fully
constraint and immobilize). And the 3rd equation indicates
that there is no feasible path from the current configurations
to any other free configurations, that means an object cannot
go out from the capture region.
As previous works about caging, Rimon and Blake for-
mulated caging condition by Morse theory, Pipattana-
somporn et al. presented a search and judge algorithms for
caging of concave polygonal and polyhedral object, .
Makita et al. tried to formulate 3 dimensional caging with
specific shapes. Wang et al. reported about application
of caging (Object Closure) to a multiple robots’ cooperative
object transfer task.
As the name “Object Closure” represents, caging has
much relation to the idea of “Closure”, and we can ac-
quire many good concept or guide-line from papers about
Closure, . For example Nguyen discusses to realize
“Force Closure”. In the paper he also analyzed the
effectiveness of a soft finger placed at a corner or edges
of an object, and referred significance of gravity as if it’s a
virtual contact point.
However those papers presented good analysis to realize
caging, but they supposed no restriction to the Cfree.obj. That
is, in the previous caging definition an object in a caging
condition can take any position and posture in the capture
region. As a general algorithm analysis previous works is
sufficient, but it is not feasible to apply the algorithm to a
real object manipulation where the position or posture should
be kept within a required conditions. For instance a coffee
cup can easily be in a caging condition by inserting robot
fingers into a ring handle, but for transferring a coffer cup
with contents, it is essential to keep horizontal posture of
the cup. As a second example, when a robot stores an object
(book or container case) to a shelf, direction and posture of
the placed object determine the quality of storing (Book title
should come to front side).
To summarize, applying caging manipulation to a posture
significant object like our target container, following two
conditions must be satisfied. (1) Position and posture must be
within required specification in its caging capture region. (2)
Especially gravity direction may have strict constraint and
must be specially cared.
In the next section, a caging framework that can overcome
two conditions will be indicated.
Cfree.obj∩ Cfree.inf= ∅
Where Cfree.obj is a 6 dimensional configuration space of
D. Geometrical caging manipulation
As a caging framework which can satisfy the above two
conditions, we propose a Geometric Object Closure (GOC).
Required conditions of GOC are as follows.
Required conditions of GOC
1) An object can move inside a defined capture
region. That means, all effect points don’t need
consistent contact with the object. Such point will
be called an effect candidate point.
2) Friction support is not counted as an effect can-
didate point because of its uncertainty.
3) Gravity direction and non-gravity direction (hori-
zontal) should be discussed in different processes.
In the process of gravity direction consideration,
gravity itself can be assumed one effect candidate
4) For discussion of displacement in 3 different
directions (Both positive and negative), one effect
candidate point must act (contact) against the
displacement within the capture region.
5) For discussion of rotation in 3 different directions
(Both positive and negative), two effect candidate
points must act (contact) against the rotation
within the capture region.
condition rejects to use frictional force for constraining an
object motion. Frictional constraint is used some application
like fork lift (Fig.4) and so on. But in these application, some
observer or sensors should be installed to monitor the object
motion, because transporter acceleration may exceed the fric-
tional constraint. In our research, simplification of an object
hand-over task is a policy to realize robust performance, but
adoption of such observer or sensors is against the policy.
The 3rd condition is based on the general significance of
supporting gravity load. The 4th and 5th conditions indicate
the caging state in a view of geometrical constraint. That
means one reactive force is essential to resist the object
displacement in a direction and one facing pair of effect
candidate points is essential to resist the moment of object
rotation. To correspond the object degree of freedom and
constraint conditions, GOC refers Morrow’s representation
in his “Primitive task” categorization.
GOC has two advantages in its deliberation process.
(1) Easy drawing process: However set-theoretical definition
(Equ. (1)) can define a caging condition very clearly, it
is not easy to apply a real application. But GOC is very
simple and easy to check while drawing (especially in CAD).
(2) Small calculation effort: As described in Wang paper,
judgment of caging state (or not) isn’t easy for arbitrary
shape objects. But GOC can be confirmed quickly, because
the object position and posture is restricted preliminarily
as a capture region, so the number of calculation target
configurations is smaller than the one of general caging.
Surely Pipattanasompornproposed a high speed caging judge
algorithm about concave polygonal object.
Detailed description of each conditions are as follows. 2nd
On the other hand, it’s not feasible to apply GOC frame-
work to general purpose human-like robot hand, so exchange
or transformation to a suitable end effecter is essential.
E. State transition between two caging conditions.
As shown in Fig.5, state transition between two different
caging conditions must avoid capture region mismatch. That
is, some work is essential in a transition from a large capture
region caging to a small capture region caging like Fig.5(B).
There are two kinds of solutions to the problem.
1) Position and posture control by additional actuators
2) Adoption of quasi-compliant mechanism whose dis-
placement range changes by its configuration.
An example of such quasi-compliant mechanism is a crane
winch mechanism (Fig.6). In the system, when the wire is
expanded the hung object can move freely, but when wound
up the object motion is restricted by a guide structure.
Because the 1st solution need sophisticated recognition
and control procedure and those make the system very
complex, we selected the 2nd solution which can omit
additional actuators and sensors.
Transition will be done naturally Some work is essential for transition
Transition TransitionTransition Transition
Fig. 5. Capture region mismatch between two caging conditions
Fig. 6.Margin size variation in a quasi-compliant winch mechanism
III. ROBOT DESIGN
Firstly this section explains system configuration for the
container hand-over task. Secondly overview of each com-
positional element will be described.
A. System configuration of container hand-over task
Fig.7 shows the system configuration. As described in
the section I, the hand-over task is performed between
the container transfer robot and the home-use automated
storage/retrieval system. The container transfer robot is com-
posed of two components; ceiling mobile component and
container handling component.
Container transfer robot
(Ceiling mobile component)
Container transfer robot
Fig. 7.System configuration of hand-over task
B. Overview of i-Container
Fig.8 shows abstract of i-Container. Functions in detail are
described in our previous paper, so this section explains
geometrical structure of i-Container which is necessary for
caging discussion. i-Container has two pin-connection holes
for the container transfer robot at top corners. The pin-
connection holes are implemented with taper guide, ac-
cordingly if connection pins have compliant mechanism the
connection motion can be performed even if there is 10[mm]
positioning misalignment between the pin and hole. Besides
i-Container has fork insertion space at the bottom and has a
hook chase at the center bottom for back and forth handling.
Pin hole with
taper guide for
robot grasping *2
view of pin
Fig. 8.Overview of intelligent container (Class A)
C. Overview of container transfer robot : ceiling mobile
Ceiling mobile component utilize permanent magnet in-
ductive traction method as shown in Fig.9. In the method
top and bottom robot components are bound by powerful
permanent magnets, the upper robot is a differential wheel
robot and moves the upper side permanent magnet, the
bottom actuation robot is navigated by the upper side robot
locomotion. As you can estimate, the upper robot motion is
in non-holonomic constraint, so heuristic approach is neces-
sary for accurate position and posture control. In addition,
because upper side of the ceiling plate is covered by 2D code
(QR code) matrix, the upper robot can estimate its position
Fig. 9. Permanent magnet inductive traction method
and posture by reading 2D code. The accuracy of position
estimation is under 0.33[mm] and posture estimation is under
0.30[deg] in standard deviation.
D. Overview of container transfer robot: container handling
Fig.10 shows overview of handling component. Container
handling component is composed of expansion and contrac-
tion component (Fig.10 right-top), crane winding component
and container manipulation component (Fig.10 bottom). Fea-
tures of the manipulation component are as follows. (1)
Crank connection pin can realize a robust container handling
motion only by inserting the connection pin into holes on
i-Containers. (2) The manipulation component is installed
with two horizontal compliant elements, and each element is
composed of 2-axes linear sliders and tension springs. (3) A
2-axes inclination compliant element is settled at the center
of body, so it can absorb slope of a target container. (4)
When grasping a container with certain load, each compliant
elements’ functions become low or invalid. Therefore stable
transport can be feasible.
Above features of the manipulation component actualize
robust container handling motion even if there are 10[mm]
position misalignment or 10[deg] inclination mismatch.
In Expansion and contraction component(Fig.10 right-top),
open steel belt actualized up/down lifting motion, and sliders
made of plastic rail and bended metal plate can prevent
unintended rotating and twisting motion of the steel belt.
When no load is applied, the sliders can prevent unintended
deformation of component, but if some external force (ex.
human contact) is applied to the component, these can
deform and reduce the contact force as shown in Fig.11.
By the effect of compliant mechanisms, the posture of the
manipulated container does not change so much.
Grasping mechanism (1)
Crank connection pin
Close up view of grasping mechanism.
Max stroke 1,200[mm]
springs * 4
Fig. 10.Abstract of container handling component
No force Applying
Fig. 11. Deformation of expansion & contraction component
Fig.12 shows overview of the container warehouse. The
container warehouse can store multiple i-Containers space-
efficiently, and store or retrieve motion can be performed
automatically. As a basic structure, the container warehouse
utilizes a book shelf in the market, and an expansion frame
actualizes the automated container warehouse. By installing
horizontal and vertical motions in different structures, the
size of drive mechanism is small enough not to invade our
living space. In the horizontal transporter, i-Container is
handled robustly in two different GOC caging conditions as
described in Fig.13, the GOC caging condition is realized by
i-Container body itself, a container guide plate, a fork table
structure and a lock plate.
Concretely speaking, the flat stand bar of the container and
the support structure of the horizontal transporter contacts
in surface. Therefore 2 rotation DOF of 2 axes in horizon-
tal plane and 1 displacement DOF of a vertical axis are
constrained by taking account of the effect of gravity. In
addition, while fork plate insertion motion, the round shape
container guides behave like two positioning pins and the
open front is restricted by the fork table push plate. That
means the rest 2 displacement DOF and 1 rotation DOF was
constrained. On the other hand, while lock plate constraint
state, the container is restricted by 2 orthogonal parallel
guides, this condition also constrains the rest 2 displacement
DOF and 1 rotation DOF.
The horizontal transporter recognizes its position by limit
switch and the accuracy is also under 1[mm]. Besides the
top of horizontal transporter is a bended metal fork table, the
bended region can guide the bottom stand bar of i-Container.
On the other hand a commercial linear actuator is utilized
for the vertical transporter and positioning accuracy is under
ofhome-use automated container stor-
Overview of automated container
On each shelf
Fig. 12. Abstract of home-use automated container storage/retrieval system
i-Container structureHorizontal transporter structure
Container guide plate structure
(B) During fork plate
Container side board
Major effective point
(C) Lock plate
(A) Neutral state
Container side board
Fig. 13.Caging condition of i-Container and horizontal transporter
F. Caging condition while container hand-over motion
As Fig.12 describes, in the container warehouse i-
Container is placed as its handle comes to the front side,
and manipulated by inserting fork table at the bottom. As
shown in Fig.10, the container transfer robot grasps at the top
of i-Container by connecting the crank pins. Consequently
the collaborative caging condition while container hand-over
motion can be summarized in Fig.14.
Front viewSide view
Container transfer robot
Fig. 14. Caging configuration of container hand-over motion
IV. IMPLEMENTATION OF HAND-OVER TASK
To implement the hand-over task, this section explains
some discussion points; (1) Hand-over execution place, (2)
Horizontal positioning method, (3) Possibility of caging state
transition and (4) Vertical positioning method. First of all,
abstract of hand-over task flow will shown in Fig.15. There
are two directional hand-over tasks as below.
• Retrieve hand-over task: In this task i-Container is deliv-
ered from the container warehouse to the container transfer
• Store hand-over task: In this task i-Container is delivered
from the container transfer robot to the container warehouse.
A. Hand-over execution place
However there are 3 candidates to execute the container
hand-over task, (1) On shelf plate, (2) On horizontal trans-
porter at static rail, (3) On horizontal transporter at movable