ArticlePDF Available

Abstract and Figures

The focus of this work is to analyze the behavior of an autonomous swarm, in which only the leader or a dedicated set of agents can take intelligent decisions with other agents just reacting to the information that is received by those dedicated agents, when the swarm comes across stationary or dynamic obstacles. An energy-aware information management algorithm is proposed to avoid over-sensation in order to optimize the sensing energy based on the amount of information obtained from the environment. The information that is needed from each agent is determined by the swarm’s self-awareness in the space domain, i.e., its self-localization characteristics. A swarm of drones as a multi-agent system is considered to be a distributed wireless sensor network that is able to share information inside the swarm and make decisions accordingly. The proposed algorithm reduces the power that is consumed by individual agents due to the use of ranging sensors for observing the environment for safe navigation. This is because only the leader or a dedicated set of agents will turn on their sensors and observe the environment, whereas other agents in the swarm will only be listening to their leader’s translated coordinates and the whereabouts of any detected obstacles w.r.t. the leader. Instead of systematically turning on the sensors to avoid potential collisions with moving obstacles, the follower agents themselves decide on when to turn on their sensors, resulting in further reduction of overall power consumption of the whole swarm. The simulation results show that the swarm maintains the desired formation and efficiently avoids collisions with encountered obstacles, based on the cross-referencing feedback between the swarm agents.
This content is subject to copyright.
remote sensing
Article
Energy-Efficient Navigation of an Autonomous Swarm with
Adaptive Consciousness
Jawad Naveed Yasin 1,* , Huma Mahboob 2, Mohammad-Hashem Haghbayan 1,
Muhammad Mehboob Yasin 3and Juha Plosila 1


Citation: Yasin, J.N.; Mahboob, H.;
Haghbayan, M.-H.; Yasin, M.M.;
Plosila, J. Energy-Efficient Navigation
of an Autonomous Swarm with
Adaptive Consciousness. Remote Sens.
2021,13, 1059. https://doi.org/
10.3390/rs13061059
Academic Editors: Józef Lisowski,
Kouzou Abdellah, Haitham Abu-Rub,
Piotr Szymak and Andrzej Stateczny
Received: 21 January 2021
Accepted: 5 March 2021
Published: 11 March 2021
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2021 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
1Autonomous Systems Laboratory, Department of Future Technologies, University of Turku, Vesilinnantie 5,
20500 Turku, Finland; mohhag@utu.fi (M.-H.H.); juplos@utu.fi (J.P.)
2Connected Shopping Ltd., Thetford IP24 1HP, UK; huma.mahboob@coursemerchant.com
3Department of Computer Networks, College of Computer Sciences & Information Technology,
King Faisal University, Hofuf 31982, Saudi Arabia; mmyasin@kfu.edu.sa
*Correspondence: janaya@utu.fi
Abstract:
The focus of this work is to analyze the behavior of an autonomous swarm, in which only
the leader or a dedicated set of agents can take intelligent decisions with other agents just reacting to
the information that is received by those dedicated agents, when the swarm comes across stationary
or dynamic obstacles. An energy-aware information management algorithm is proposed to avoid
over-sensation in order to optimize the sensing energy based on the amount of information obtained
from the environment. The information that is needed from each agent is determined by the swarm’s
self-awareness in the space domain, i.e., its self-localization characteristics. A swarm of drones as
a multi-agent system is considered to be a distributed wireless sensor network that is able to share
information inside the swarm and make decisions accordingly. The proposed algorithm reduces the
power that is consumed by individual agents due to the use of ranging sensors for observing the
environment for safe navigation. This is because only the leader or a dedicated set of agents will
turn on their sensors and observe the environment, whereas other agents in the swarm will only
be listening to their leader’s translated coordinates and the whereabouts of any detected obstacles
w.r.t. the leader. Instead of systematically turning on the sensors to avoid potential collisions with
moving obstacles, the follower agents themselves decide on when to turn on their sensors, resulting
in further reduction of overall power consumption of the whole swarm. The simulation results show
that the swarm maintains the desired formation and efficiently avoids collisions with encountered
obstacles, based on the cross-referencing feedback between the swarm agents.
Keywords:
autonomous swarm; multi-agent systems; energy efficient; swarm intelligence; leader
follower; collision avoidance
1. Introduction
The optimization of autonomous navigation, collision avoidance, and resource al-
location in swarms of drones (Unmanned Aerial Vehicles or UAVs) is currently one of
the major focus areas in the robotics research community [
1
]. Besides the usability of
individual UAVs, the considerable advantages of utilizing swarms of UAVs have increased
their demand in various fields, such as search and rescue, traffic monitoring, atmospheric
research, and military applications [25].
A general categorization of the agents in a swarm can be presented, as follows [68]:
Evolutionary agents, where the agents work on the fundamental theory of evolutionary
algorithms i.e., mutation, reproduction, recombination, and selection.
Cognitive agents, where the agents can take decisions, make predictions, and process
data based on their cognitive architecture.
Reactive agents, where, as the name suggests, the agents react to a signal from another
agent or to any change in the surrounding environment.
Remote Sens. 2021,13, 1059. https://doi.org/10.3390/rs13061059 https://www.mdpi.com/journal/remotesensing
Remote Sens. 2021,13, 1059 2 of 17
Flocking agents, where the agents imitate the flocking behavior, i.e., moving together,
inspired by flocks of birds or swarms of bees, for instance.
When considering the navigation of a swarm of drones, collision avoidance and
formation maintenance in the swarm are the two of the most prominent problems to
solve [
9
11
]. With the exponential increase in the use of UAVs and their integration into
many different commercial, military, and leisure applications, the need for an efficient
onboard collision avoidance system increases exponentially. Such an onboard system
enables a drone to react rapidly to encountered objects during flight [
12
,
13
]. Owing
to, e.g., onboard payload limitations, power limitations, and complications in remote
monitoring, tasks become increasingly difficult for UAVs to accomplish. The robotics
community is trying hard to counter these issues by developing new technologies to ensure
safe navigation for UAVs in various environments [1417].
Collision avoidance systems or algorithms are responsible for safely and reliably
avoiding any possible collisions amongst the agents (e.g., UAVs) themselves and between
an agent and a surrounding obstacle in the environment [
18
]. Collision avoidance algo-
rithms can be roughly classified into the following three categories [
19
]: (1) sense-and-avoid
algorithms that simplify the process by delegating the detection and avoidance activities to
individual agents/nodes and, therefore, they have short response times, are independent
of inter-node communication, and require less computational power [
20
22
]; (2) force-field
algorithms, also known as potential field methods, which use the basic concept of attrac-
tion/repulsion between the agents in the swarm and between the agents and obstacles
in the environment to guide an agent towards a destination while avoiding objects along
the path [
23
26
]; (3) optimization based algorithms, relying on geographical informa-
tion, which utilize knowledge on the sizes, shapes, and locations of obstacles to provide
near-optimal path planning solutions [2729].
Formation control can be divided into several separate tasks: navigation of the whole
formation/swarm from one point to the designated point, maintaining a certain formation
shape or orientation, splitting the formation, bringing the agents back into the original
formation, and avoiding collisions while accomplishing these tasks [
30
]. In more general
terms, a formation can be defined as the shape where the positioning of each agent within
the swarm is relative to other agents [
31
33
]. Formation maintenance algorithms can be
outlined into the following three classes [
34
,
35
]: (1) leader–follower based approaches, in
which all of the agents in the swarm follow the leader and autonomously maintain their
respective positions, w.r.t. their neighbours and the leader [
36
39
]; (2) virtual structure
based approaches, in which all of the agents of the swarm as a whole are considered to be a
single compound agent to be navigated along a given trajectory [
40
43
]; and, (3) behavior
based approaches, in which the agents select their behavior in each situation based on a
pre-determined procedure or strategy [44,45].
In this paper, we propose a strategy to reduce the processing power of individual
agents in the swarm without losing the swarm’s ability for autonomous operation. In order
to achieve this, a leader-follower based approach is adopted for maintaining the formation,
due to its relatively simple implementation, scalable nature, and reliability [
9
,
38
]. The global
leader in the formation utilizes a given global collision avoidance algorithm, which is then
used by the followers for the calculation of the relative coordinates, which are also known
as translational coordinates, w.r.t. themselves, of the detected objects in the environment,
as observed by the global leader. Furthermore, followers themselves, upon receiving the
observed coordinates of a detected obstacle from the leader, compute to decide the optimal
time for turning on their sensors for dynamic avoidance if some error has been detected
in the calculation of translational coordinates while cross-checking is performed by the
leader. In the case an error has been detected, it indicates that the obstacle is not stationary
anymore, i.e., the environment is not static and, hence, it is treated as a moving obstacle,
i.e., the environment is dynamic.
In this paper, a new strategy for reducing the overall energy consumption of a swarm
is proposed. The main idea is to remove the unnecessary power consumption due to
Remote Sens. 2021,13, 1059 3 of 17
the sensors for a portion of the swarm in cases where the information the agents per-
ceive from their neighbors is enough for individual navigation of an agent. In other
words, the neighbors of each agent are playing as a source of information for navigational
purposes. The main contribution of the proposed approach is that besides the existing
energy-efficient methods, such as movement-based and communication-based methods,
the energy consumption of the swarm can be further reduced by injecting adaptive con-
sciousness in the agents, especially in scenarios where the environment is either static or
dynamic variables in the surroundings are negligible. In such situations, the agents can
turn off their ranging sensors, translate the coordinates transmitted by their leader, and,
upon performing necessary calculations, switch to the high-conscious mode if deemed to
be necessary.
The organization of the rest of the paper is as follows. Section 2provides the motiva-
tion. Section 3provides the development of the proposed algorithm in detail.
Section 4
provides the simulation results. Finally, Section 5provides the concluding remarks, discus-
sion, and future work.
2. Motivation
The minimization of energy consumption of UAVs, or autonomous mobile agents
more generally, for the efficient utilization of their limited resources, is currently gaining
the attention of researchers. This is of high interest in scenarios where a prolonged mission
duration is desirable, such as in large warehouses, underwater exploration, patrolling,
and guarding [
38
]. However, in the literature, most of the energy-efficiency related work
focuses on countering the adverse effect of external influences [
46
], route optimization [
47
],
recharging optimization [
48
], altitude based navigation [
49
], or exploiting the direction of
wind to extend mission duration [50].
In a multi-agent system, it is a norm to have every single agent making more or less
intelligent decisions by utilizing the onboard sensor system for observing the surroundings.
The active sensors and onboard processing of sensory data consume power and reduce
battery run-time, which results in decreased mission life on a single charge. However,
in static environments, all of the agents in the swarm do not necessarily have to either keep
their onboard sensors turned on continuously or take decisions intelligently. If intelligent
decision making and sensor usage are restricted based on some constraints, depending on
the environment and surroundings, a major portion of power consumption due to sensors
and related data processing can be reduced, as shown in Figure 1. The approach that is
presented in [
17
] proposes a translational coordinates scheme in which the leader agent
has its sensor turned on all the time, while the follower agents turn on their sensors as soon
as the leader detects any static obstacle in its surroundings. Indeed, for such environments,
if only the leader of the swarm remains at a high-conscious mode continuously and
performs all of the necessary required calculations (Figure 1a), the system’s sensor related
power consumption can be reduced considerably, and the potential run-time of the mission
is thereby increased. Furthermore, the adaptive capability of the individual follower agents
to dynamically switch between the low-conscious and autonomous high-conscious modes,
depending on the situation at hand (Figure 1b), guarantees that the system’s degree of
autonomy is not compromised.
In this paper, based on the principle that is described above, an approach is developed
to reduce the overall power consumption of the system by intentionally deactivating and
reactivating the sensors at run-time. Basing our proposed technique on reactive agents [
6
]
and a control methodology for formation based on 1-N leader-follower approach (1-leader
and N-followers) [
38
], the leader constantly observes the environment and informs its
followers of the coordinates of any detected obstacles in the vicinity. In case there are no
obstacles in the vicinity of the leader, it only broadcasts its own coordinates. The follower
agents, after translating the coordinates according to their own positions, decide whether
continuing the same trajectory is safe or if they are required to deviate from their path.
However, in the case a dynamic obstacle is detected by the leader agent, the coordinates
Remote Sens. 2021,13, 1059 4 of 17
are broadcasted in a similar manner to the followers, who, then upon performing the
necessary calculations, realize the direction of movement of the obstacle and its speed.
After this, based on their own trajectory and the obstacle’s trajectory, it is calculated if
continuing on the same trajectory will lead to a potential collision, in which case the
respective agent switches from the low-conscious mode to the autonomous high-conscious
mode, as shown in Figure 1. After successfully avoiding the collision or bypassing the
obstacle, the follower agent switches back again to the low-conscious mode. Individual
agents in the swarm utilize the collision avoidance technique developed and presented
in [
9
] for avoiding collisions in the high-conscious mode. The main motivation behind
this approach is not only the overall power cost reduction of the swarm, but also the
capability of the agents to autonomously decide on switching between the low-conscious
and high-conscious modes, enabling situation-aware optimization of power consumption,
due to ranging sensors, at run-time. A simplified approach has been chosen for validation
of the proposed approach by limiting the depth of the swarm to two agents, i.e., ”N” = 2
agents in the 1-N leader–follower model. It is important to note here that the agents in
the Follower_Mode are assumed to utilize the on-board IMU (Inertial Measurement Unit)
and GPS (Global Positioning System) for localization purposes, i.e., obtaining their own
positions (localization methodologies are not in the scope of this work. However, while
in low-conscious mode, the possible error in obtaining position vectors due to the IMU
drift over time and also the probable GPS signal loss, is also handled by the algorithm, as
the cross-checking of the translated coordinates will return false readings and the follower
agent will go into the high-conscious mode temporarily ).
(a)Scene 1 (b)Scene 2
Figure 1.
Swarm in a static environment with only one dynamic obstacle. (
a
) the initial scene
where the leader (encircled by light green) has its ranging sensors turned on, performs all the
processing, and transferring the information about the surroundings to its followers, (
b
) the Follower
2, upon performing necessary calculations, realizes that the dynamic obstacle is potentially moving
in its path and turns on its sensors to be able to perform collision avoidance actively. In both scenes,
Follower 1 does not turn its sensors on, since the dynamic obstacle is moving in the other direction
and does not pose a potential collision threat. For bypassing the static obstacles, Follower 1 translates
the received coordinates to generate escape routes.
3. Proposed Approach
In this section, we describe the proposed method in more detail. The strategy is to
combine the navigation and object detection with coordinate calculation and adaptive
autonomous modes in order to facilitate the process of autonomous swarm navigation
using translational coordinates. The proposed top-level algorithm to accomplish this
(Algorithm 1) is composed of two partial feedback-based algorithms: one for managing
Remote Sens. 2021,13, 1059 5 of 17
the followers and the other for managing the adaptive autonomous mode in the presence
of obstacles. The adaptive autonomous mode calls collision avoidance from within if
deemed to be necessary.
In the Follower_Function module, drones receive the coordinates of their leader and
of the obstacles, as seen and transmitted by the leader, and then perform the necessary
calculations to translate the received coordinates according to their own respective positions.
Meanwhile, if there is no feedback from the leader and the defined timeout is exceeded,
it is assumed that the leader is not alive/reachable, i.e., the communication link with the
leader is lost. In such a case, the drone in question temporarily declares itself as its own
leader as a fail-safe mechanism and resumes navigation by turning its sensors on.
The environment is declared to be dynamic, if the leader detects one or more moving
obstacles, or if the cross-check of the translational coordinates indicates a mismatch. The fol-
lower, in this case, starts calculating the obstacle’s velocity to determine the apparent point
of impact. Upon estimating the point of impact, the follower node/drone itself decides
when it needs to turn on its sensor(s) in order to perform active collision avoidance for
safe maneuvering.
Algorithm 1 Leader: Navigation & Object Detection
agent Leader()
2: while True do
if obstacle detected then
4: Dobstacle,Aobstacle Obstacle’s distance and angle calculation;
end if
6: for all i do
Follower(i).Transfer_Coordinates(dt f ,df f (i),tsuccess );
8: if tsuccess then
cal.re f .coords Reverse cross-check follower’s received
coordinates(df f (i));
10: cal.re f .angle Reverse cross-check follower’s received angle(df f (i));
coords,angle =re f .coords cal.re f .coords,re f .angle cal.re f .angle;
12: if |coords|>Threshol dcOR |angle|>Thresholdathen
Follower(i).Set-Dynamic; .Setting remote Dynamic flag
14: end if
end if
16: end for
Collision_Avoidance();
18: end while
end agent .end agent leader
3.1. Agent Leader
Algorithm 1provides the general pseudo-code for the global leader. This top-level
algorithm is executed utilizing the on-board processing units by all agents locally. The as-
signment of IDs to all agents is assumed to be achieved before starting the mission. The al-
gorithm is initialized by first setting up the necessary flags. Subsequently, the leader
agent starts scanning for any obstacles in the vicinity and then calculates the distance
and angles at which the detected obstacle(s) lie (Lines 3–4). After this, the leader starts
sending its coordinates to its respective follower(s), including the detected obstacle’s (if
any) angle and distance (Lines 6–7). Here, the symbols,
dt f
and
df f
, are used to simplify
the notations and contain the distance of the obstacle(s), the angles at they lie, the agents’
own parameters, such as coordinates and heading direction. After receiving the coordi-
nates from its follower(s), i.e., the transfer is successful (
tsuccess
), the leader cross-checks
the distances and angles calculated by the follower (Lines 8–10). After cross-checking,
the environment is declared to be dynamic if the absolute value of the received angles
or coordinates, i.e.,
coords
,
angl e
, is more than a defined threshold level, i.e.,
Thresholdc
Remote Sens. 2021,13, 1059 6 of 17
error tolerance for coordinates and
Thresholda
error tolerance for calculated angle (Lines
12–14). The leader agent then calls Collision_Avoidance to bypass the obstacle(s) successfully
(Line 17).
3.2. Agent Follower
Algorithm 2provides the general pseudo-code for the follower agents. The algorithm
starts by declaring its global procedures that are also utilized by the leader and, furthermore,
setting up the necessary flags (Lines 2–3). Afterwards, the follower agent(s) check if the
leader is alive, i.e., they are getting constant feedback from the leader (Lead_is_Alive ==
True), or if they are in the follower mode, i.e., Follower_Mode == True. If either Lead_is_Alive
or Follower_Mode is False, the respective agent performs obstacle detection actively and then
calculates the distance and angle at which the obstacle is detected (Lines 5–6). In the case of
constant feedback from the leader, the respective follower agents check if the environment
has been declared as dynamic by the leader or not. If the environment is not declared
as dynamic, the agent calls Follower_Function for navigational purposes. On the contrary,
if the environment has been declared to be dynamic, Adaptive_Mode_Function (Lines 8–11).
Finally, the Coordinates_Received flag is reset to False, for next iteration (Line 13).
Algorithm 2 Follower: Navigation & Object Detection
agent Follower()
2:
global procedure Transfer_coordinates, Set-Dynamic;
.
declaration of of follower’s
global procedures Leader will call
Lead_is_Alive,Fol low_Mode,Dynamic,coordinates_received = True, True, False,
False;
4: while True do
if (!Lead_is_Alive OR !Follower_Mode) AND obstacle detected then
6: Dobstacle,Aobstacle Obstacle’s distance and angle calculation;
end if
8: if !Dynamic AND Lead_is_Alive then
Follower_Function();
10: else
Adaptive_Mode_Function();
12: end if
coordinates_received = False;
14: end while
end agent .end agent follower
3.3. Set-Dynamic
Algorithm 3shows the procedure that is called by the leader to set the dynamic flag
for its followers to True.
Algorithm 3 Set-Dynamic
procedure
SET-DYNAMIC()
.
this procedure is called remotely to set Dynamic to True
for Follower(i)
2: Dynamic = True;
end procedure
3.4. Transfer Function
Algorithm 4specifies the pseudo-code for Transfer Function. The algorithm starts
by setting the
tsuccess
, i.e., the transfer success, flag based on the
coordinates_received
and
randomizing the True/False to simulate the possibility of transfer failure (Line 2). Sub-
sequently, if
tsuccess
is True, the data sent by the leader are utilized to compute the trans-
lated coordinates by the follower (
df f
). If there was a constant feedback from the leader,
the node/follower uses the translational coordinates of the obstacle, as observed by the
Remote Sens. 2021,13, 1059 7 of 17
leader and based on its own coordinates, calculates the location of the obstacle according
to its own position, as shown in Figure 2variables description given in Table 1). These
translated coordinates that are calculated by the follower are then utilized by the leader,
as data from the follower,
df f
, (Algorithm 1, Lines 17–19) for cross-checking purposes.
The coordinates_received flag is then set to True (Lines 3–6).
Algorithm 4 Transfer Function
procedure TRANSFER_COORDINATES(dt f ,df f )
2: tsuccess
= !
coordinates_received
AND random select(True, False);
.
modeling transfer
failure possibility
if tsuccess then
4: receiveddata =dt f ;
df f = calculate translational coordinates(dt f ); .Follower’s data for Leader
6: coordinates_received = True;
Lead_is_Alive = True; .if Lead_is_Alive is False, turn it to True
8: end if
end procedure
Figure 2. Distance and direction calculation of the detected obstacle.
Table 1. Description of Variables from Figure 2.
Variables Description
DL distance of the obstacle’s left
DR and right edges from leader
D1Ltranslational calculated distance of
D1Rright and left edges of the obstacle
D2Lfrom follower 1 and follower 2,
D2Rrespectively, as observed by leader
d1follower 1 and follower 2’s distance
d2from leader respectively
θRangle of detected right and left edges
θLfrom leader respectively
θ1L leader’s angle from follower 1 and
θ2L follower 2, respectively
ϕF1R angles of right and left edges as
ϕF1L calculated by follower 1 and
ϕF2R follower 2, respectively
ϕF2L
Remote Sens. 2021,13, 1059 8 of 17
3.5. Follower Function: Coordinate Calculation
Algorithm 5specifies the Follower_Function module, where the follower node waits
until either the defined Timeout is reached or the coordinates are received (Line 2). Sub-
seuently, if the coordinates_received is True, i.e., coordinates are received by the follower,
the reference coordinates that the follower should navigate to, i.e., ref.coords, are updated
w.r.t.
df f
(Lines 3–4). Moreover, if the ranging sensors are on/activated previously, they are
turned off, as the agent is now in the follower mode (lines 5–6). However, if the coordinates
are not received until the Timeout, the Lead_is_Alive is set to False, ranging sensors are
turned on for actively observing the environment and perform obstacle avoidance, and to
make itself its own leader, the node sets the ref.coords as its own coordinates, i.e., self.coords,
to start navigating towards the destination (Lines 8–11).
Algorithm 5 Follower Function
procedure FOLLOWER_FU NC TI ON()
2: wait until Timeout OR coordinates_received;
if coordinates_received then
4: re f
.
coords df f
;
.
Update the reference coordinates to translational coordinates
if Sensors are ON then
6: Turn off the Sensors;
end if
8: else
Lead_is_Alive =False;
10: Turn on the Sensors;
re f .coords =s el f .coords;
12: end if
end procedure
The timeout signal is only checked if the node in question is not the leader itself. Every
other node constantly checks whether its respective leader’s transmitted signals are being
constantly received. If the node has not received the coordinates sent by its leader by the
timeout, it turns on its sensors for active collision avoidance maneuvering by declaring
itself as its own leader.
3.6. Adaptive Mode Function
This module, Algorithm 6, is called by the leader if it detects obstacles in the vicinity.
Or, in the case the environment is already as dynamic, then the nodes in Follow_Mode call
this module locally. As soon as this module is called, it is checked whether the node in
question is the global leader or is it one of the followers (Line 2). The default or initial
value of Follow_Mode for all of the follower nodes is True. If the node is the follower, using
previous translational readings, the obstacle’s velocity is approximated (Line 3) using
Equations (1)–(3), and a visual illustration is shown in Figure 3.
Figure 3. Illustration of calculation of movement of obstacle.
Remote Sens. 2021,13, 1059 9 of 17
Algorithm 6 Adaptive Mode Function
procedure ADAPTIVE_MODE_FUNCTION()
2: if Follow_Mode then
Vobs Calculate obstacle’s velocity from previous readings(dt f );
4: Dimp Calculate the distance to impact;
if Dimp <Detection Range then
6: Turn on the sensors;
Follow_Mode =False;
8: end if
end if
10: Collision avoidance ();
if Dobstacle NOT in Detection Range then
12: Dynamic =False;
if Leader != Sel f AND Lead_is_Alive then
14: Follow_Mode =True;
Turn off the sensors;
16: end if
end if
18: end procedure
Three scenarios are possible based on
Vobs
: (1) if
Vobs <
0, i.e., positive, this means that
the obstacle is going away from the node; (2) if
Vobs
= 0, this means that the obstacle is not
moving, i.e., stationary; (3) if
Vobs <
0, this means that the obstacle is travelling towards
the node. Based on these readings, the distance to the potential impact/collision (
Dimp
) is
calculated (Line 4). The distance travelled by the node after t1seconds is computed by:
dn=vn(t1t0)(1)
where
dn
and
vn
are the distances travelled by the node and velocity of the node, respec-
tively. The distance covered by the obstacle and its velocity can be determined by (2) and
(3), respectively:
dobs =dodnd1(2)
vobs =dobs/t(3)
where
dobs
is the distance travelled by the obstacle,
vobs
velocity of the obstacle,
dn
the
distance travelled by the node,
d1
the distance between the obstacle and the node at time
t1, and dois the distance between the obstacle and node at the previous time t0.
If the distance that is travelled by the obstacle is zero, i.e.,
dobs
= 0, it means that the
obstacle is stationary. However, if the distance covered by the obstacle after
t1
, i.e.,
d1
is
less than the distance between the node and the obstacle (
do
), in that case, the obstacle is
moving towards the node (as shown in Figure 3). Otherwise, the obstacle is moving away
from the node. Based on the movement of the node and the obstacle, the apparent distance
to impact is calculated and, if that distance to impact is less than the detection range
of the on-board sensor system, the node turns on its sensors and comes out of follower
mode to perform active collision avoidance (Lines 5–7). The collision avoidance module is
then invoked to constantly monitor the environment (line 10). After successful collision
avoidance, the status of the surroundings is changed back to static (Lines 11–12). The node
turns off its sensors and goes back into Follow_Mode and the control is returned to the main
module (Lines 13–16).
3.7. Collision Avoidance
Collision avoidance, with the pseudo-code in Algorithm 7, is invoked when the
detected obstacle gets critically close. It is then checked whether there is only one obstacle
in the vicinity or there are multiple obstacles (Line 3). In case, more than one obstacle is
detected, the gap between the obstacles is calculated (Line 4). Based on the calculation,
Remote Sens. 2021,13, 1059 10 of 17
the algorithm takes the following actions: if the
gap
is more than the
sa f e_dist
, i.e., defined
based on the dimensions of the UAV, then the UAV is aligned to pass through the gap
available between the obstacles (Lines 5–6); otherwise, the obstacles are treated as a single
obstacle and the UAV is rerouted accordingly to navigate around them in order to avoid
collisions (Lines 7–9). However, if only one obstacle was detected in the first place, the path
planning is done accordingly to bypass the obstacle while keeping the deviation to a
minimum (Lines 11–13). For performing successful path planning and aligning, we utilized
and implemented the technique that is presented in [
9
]. The algorithm then recalculates
the obstacle’s distance and updates it. Based on the updated value, the control returned
back to the Adaptive_Mode_Function.
Algorithm 7 Collision Avoidance
procedure COLLISION_AVOIDANCE()
2: while Dobstacle <Detection Range do
if detected obstacles > 1 then
4: gap calculate the gap between obstacles;
if gap >sa f e _dist then
6: path planning(edges); .UAV is aligned to pass through the obstacles
else
8: path_plan single obstacle;
path planning(path_plan); .considered as single obstacle
10: end if
else
12: path_plan single obstacle;
path planning(path_plan);
14: end if
Dobstacle Update the obstacle distance;
16: end while
end procedure
4. Simulation Results
The area used for the basis of the simulation was defined as 700 m
×
500 m two-
dimensional XY-plane, i.e., all of the objects are considered to be at the same altitude.
The number of agents is set to three and the agents are already in the defined V-shaped
formation at the start of the mission. It is important to note that, in the performed ex-
periments, the leader agent has more computational and power resources to be able to
perform the leadership tasks. The leader was equipped with Nvidia Jetson TX2, which is
a power-efficient embedded system with the capability of operating between
0.5–2 GHz.
In order to be more power efficient, we set the operating frequency to 0.5 GHz, at which
the average power consumption is 4.5 Watts. The followers were equipped with Raspberry
Pi 3B, which consumes around 2.4 Watts in high-conscious mode and 1.4 Watts in the idle
mode. Velodyne Puck LITE was used for the generation of the data, which was then used
in the simulation platform for visualization purposes, simulating the sensor, generating
the obstacles, and subsequently to verify the proposed algorithm. Puck LITE has a 360
field of view horizontally, 30
vertically, generates 300,000 points/second, and has an
accuracy of
±
3 cm and range of 100 m [
51
]. For communication between the agents, due to
its longer transmission range (100 m indoor/urban environment), data rate of 250 kbps,
and possibility of large number of devices to be connected, communication-based power
consumption is evaluated based on Legacy Digi XBee-Pro S1 802.15.4 consumption, i.e., in
Transmitting Mode = 710 mW, in Receiving Mode = 182 mW [52].
The following assumptions and initial conditions are used in this study:
1. all of the agents are travelling with constant ground speeds;
2.
communication link between the agents is setup ideally and with no information loss;
and,
3. agents utilize the on-board localization techniques in order to obtain their positions.
Remote Sens. 2021,13, 1059 11 of 17
The LiDAR (Light Detection and Ranging) data [
53
] is shown at different time intervals
and from different angles shown in Figure 4. Figure 4b–d show the obstacles when they
are in close vicinity and in the detection range. These figures show the position of the
LiDAR sensor equipped drone as the blue, red, and green interactive marker. It is to be
noted that the developed algorithm takes the point cloud that is captured by the LiDAR
based on the defined constraints to reduce the complexity of the algorithm. Because its
running on a resource-constrained system it is crucial to discard unnecessary point clouds,
i.e., point clouds that are not in the field of view (FoV). The FoV is defined in the proposed
algorithm to be
±30.
Only the obstacles within this FoV are considered to be posing
a potential collision threat to the agents and, therefore, any obstacles outside this are
discarded. The detected obstacles are contoured by light blue rectangles.
(a) (b)
(c) (d)
Figure 4.
Light Detection and Ranging (LiDAR) point set snapshots at different intervals: starting
point of simulation and when the obstacles are visible. (
a
) LiDAR data 1: at the start of the simulation.
(
b
) LiDAR data 2: obstacles are starting to enter the detection range. (
c
) LiDAR data 3: obstacles are
detected and in close proximity. (d) LiDAR data 4: bypassing the obstacles.
Figure 5shows the simulation results for a static environment when the obstacles are
stationary. In the V-shaped formation that is shown in the figure, agent 1 (leader) is shown
as a blue circle, which has its ranging sensor (shown in Figure 4) on always-on mode, while
red and green followers, i.e., agent 2 and agent 3 respectively, are following the leader
by translating the leaders’ transmitted coordinates as shown in Figure 2. In Figure
5b,
the obstacle has entered the detection range of agent 1 but since there is no danger of
collision, agent 1 continues its trajectory as shown in Figure 5c. Similarly, agent 2 is
following agent 1 while utilizing the translational coordinates, and as there is no collision
risk found upon calculations it continues to function in the Follower_Function. Whereas,
as shown in
Figure 5b,
the translational coordinates calculations performed by
agent 3
clearly indicate the potential collision if the same trajectory is continued and, therefore, it
diverts from its original trajectory by performing offline collision avoidance to avoid the
obstacle, as can be seen from the traces of the agents in Figure 5c.
Remote Sens. 2021,13, 1059 12 of 17
(a) (b)
(c) (d)
Figure 5.
Simulation results at different time intervals, showing the behavior of the agents while
going through obstacles. (
a
) Mission start with agents in formation, with leader in blue, agent 2 in red,
and agent 3 in green. (
b
) Obstacle in range, agent 3 deviates by performing necessary calculations.
(
c
) Traces of movement while going through the obstacles. (
d
) Agent 3 comes back to its intended
position in the formation when there is no obstacle.
Agent 2 maintains its trajectory w.r.t. the leader without deviating from the original
trajectory even while going through the obstacles and second obstacle being close to it, as
can be seen from Figure 5c,d. This shows the effectiveness of utilizing the translational
coordinates as the obstacle was out of the collision radius as also indicated by the performed
calculations by the agent. Whereas, agent 3 maintains the pre-defined minimum distance
from agent 1 after avoiding the obstacle, which is due to the third obstacle in the bottom.
In Figure 5d, the destination point was moved at run-time to demonstrate the reformation
process and, as can been seen, agent 3 starts going back to the position in the formation, it
was supposed to be, as soon as it moves away from the obstacle in the bottom.
Figure 6illustrates the simulation results for dynamic environment scenarios, where:
Case 1: only one obstacle is moving, as shown in Figure 6a–c. In this case scenario,
the obstacle is moving in the direction of the swarm. The obstacle, as observed by agent 1,
will intercept the swarm and a potential collision is evident if agent 3 continues its trajectory,
as can be seen from Figure 6b. Therefore, after performing the necessary calculations, agent
3 turns on its own ranging sensor and performs collision avoidance actively. After avoiding
the obstacle, agent 3 turns off its sensors and starts following the leader once again based
on translational coordinates from the leader. Figure 7a shows the distance maintained
by the agents throughout the simulation; Case 2: two obstacles are moving, as shown in
Figures 6d–f. In this case scenario, the first moving obstacle only disturbs agent 3, as in
Case 1. However, after performing the translational coordinate calculations, as soon as
agent 2 observes that the second obstacle is moving towards their direction and a collision
may be inevitable, it turns on its ranging sensor to actively avoid the collision and divert
from its trajectory if necessary (as shown in Figure 6f). After successful avoidance, when
there are no moving obstacles in the vicinity, the agents turn off their sensors and switch
back to the Follower_Function. Figure 7b shows the distance maintained by the agents
throughout the simulation.
Remote Sens. 2021,13, 1059 13 of 17
(a) (b) (c)
(d) (e) (f)
Figure 6.
The results shown from the simulations of two case scenarios showing the traces of move-
ment and deviations due to moving obstacles, with one moving obstacle and two moving obstacles.
(
a
) Mission start. (
b
) Moving obstacle observed. (
c
) Overall trace of movement. (
d
) Mission start. (
e
)
Movement of Obstacle 1 observed. (
f
) Movement obstacle 2 observed and traces of the deviated path.
(a)With one moving obstacle
0
20
40
60
80
050 100 150 200 250 300 350 400
Distance (m)
Time (s)
distance 21 distance 31
distance 32 Safe Distance
(b)With two moving obstacles
Figure 7. Distance maintained by each agent from its neighbour.
We report and compare the power consumed by our setup based on the experiment to
analyze the efficiency of the proposed algorithm. The battery used in the setup for calcula-
tion of these attributes was 5000 mAh and, in typical conditions, the power consumption
of one Velodyne Puck LITE is 8 Watts and operating voltage is 9 volts. Based on these
values and the tracking the amount of time, each agent had its sensor turned on during
the simulation with and without our proposed methodology, as shown in
Figure 8
. We
placed these calculated values side-by-side with the reference algorithm [17]. The normal
setup with all agents running their sensors always-on mode will consume about 3000 mWh,
which is calculated as N
agent=1
Eagent =Ps,agent tα,agent (4)
where the summation is over all the agents in the system,
Eagent
denotes the energy con-
sumption of an agent,
Ps,agent
is the power consumption of the agent’s sensor, and
tα,agent
is the total time for which the agent’s sensor remains active. In the instant case, there are
three agents, the power consumption of the sensor is 8000 mW and the sensors remain
active all the time, i.e., 450 s,
Eagent
comes out to be 1000 mWh and the total consumption
for three agents will be 3000 mWh.
Figure 8. Energy Consumption by all nodes (mWh).
Remote Sens. 2021,13, 1059 14 of 17
Similarly, the setup that was used in [
17
] consumed 1595 mWh, as in that setup
the follower agents turn on the sensors as soon as the leader detects any static obstacle.
However, while utilizing the proposed modified algorithm, the energy consumption for
the same setup can be reduced by about 600 mWh to 1000 mWh. As in the proposed
methodology, only agent 1 (the leader) had its sensor always-on, while agents 2 and 3
never had to turn on their sensors, as they only translated the broadcasted coordinates by
the leader and followed. Subsequently, the power consumption due to ranging sensors
in static environments can be further reduced by another 40% mark approximately. Addi-
tionally, that can be further utilized when the environment has some dynamic variables
involved and, hence, increasing the overall mission life on a single charge. When testing the
proposed methodology for the environment with some dynamic variables, the following
results were observed: (1) Dynamic S1: is the experiment with swarm consisting of three
agents and obstacle 1 moving. This only affected agent 3, and agent 3 turned on its sensor
to perform collision avoidance for 37s and turned off the sensor after successful avoidance
to go back into the Follower_Function resulting in the overall energy consumption due to
sensor usage by all nodes to 1083 mWh; (2) Dynamic S2: is the experiment performed with
swarm that consists of three agents and obstacle 1 and 2 are both moving. This affects both
agents 2 and 3. Agent 3 had to turn on its sensor for 37 s, while agent 2 had its sensor
turned on for 68 s, which resulted in the overall energy consumption due to sensor usage
by all nodes to 1234 mWh; and, (3) Dynamic S3: is the experiment performed with swarm
consisting of five agents and both obstacles 1 and 2 are dynamic. This affects all of the
agents. Agents 2 and 3 remain at high-conscious mode for the same amount of time as in
“Dynamic S2” case, whereas agents 5 and 4 remain at high-conscious mode for 57 s and
65 s, respectively. Resulting in the overall energy consumption due to sensors usage to
1506 mWh.
Based on the incremental consumption due to dynamic obstacles in the environment,
Figure 9shows an estimation of the relationship between the dynamicity of the environ-
ment and the energy consumption of the swarm. It is important to note that the energy
consumption due to the sensor(s) usage depends on the duration that each agent stays at
the high-conscious mode. It can be concluded that the energy consumption of the swarm is
independent of the number of dynamic obstacles; however, it is dependent on the dura-
tion that each agent is affected by the dynamic obstacles or has to stay at high-conscious
mode, i.e., their sensor(s) turned on. In the figure,
A=
1 represents one agent operating
in high-conscious mode,
A=
2 represents two agents operating in the high-conscious
mode,
A=
3 shows the consumption of the swarm when three agents are operating in the
high-conscious mode, and
A=
5 shows the consumption of the swarm when five agents
are operating in the high-conscious mode.
Figure 9.
Energy consumption of the swarm based on the average duration that the agents stay at
high-conscious mode.
5. Conclusions and Future Work
We developed an algorithm for a pre-defined formation of multi-agents with adaptive
ability for static and dynamic environments to reduce the power consumption due to
sensors usage. Under normal conditions in a static environment, one dedicated agent
Remote Sens. 2021,13, 1059 15 of 17
(leader) can see its surroundings with the sensors turned on, while other agents (followers)
blindly follow by translating the received coordinates from the leader. However, according
to the demand of the environment, the follower(s) have the ability to decide to turn
on their ranging sensors for safe maneuver, after performing the necessary calculations.
The proposed methodology of the effectiveness of translational coordinates and the proof of
concept was verified in the simulation environment. It is evident from the simulation results
that the proposed method is reliable in static environments, as well as in the environment
with some dynamic variables. Furthermore, it helps in reducing the power consumption
due to the usage of sensors over time. The proposed method, in the considered test case
and static environment, helped in reducing the power consumed by the sensors by about
40% when compared to the reference algorithm. In general, it is obvious that this power
saving in dynamic environments definitely depends on the structure of the environment
and its dynamicity. However, in environments with some dynamic variables in our test
scenarios, the proposed methodology still proved to be effective, as utilizing it kept the
power consumption about 50% less than if the sensors were used in a continuous mode.
In the future, we plan to further extend this work by increasing the depth of the
swarm, i.e., the number of followers in the swarm, in order to further investigate the
information exchange between the agents by taking the possible communication delays
into consideration. Furthermore, the power consumption due to wireless communication
and the effect of processing the information and coordinates on the battery life will also
be analyzed. Because the leader undertakes all of the necessary computations, leading
to faster draining of its battery and resultant shortening of overall life time of a mission,
it will be interesting to analyze the effectiveness of an election-based leadership solution
and explore different options for optimally swapping the leadership role amongst the
followers, e.g., swap the places by selecting the follower with highest remaining battery
life. Further, the effect of such improvements on the overall mission life needs to be studied.
Additionally, finally moving onto testing the proposed approach in real-time initially under
static environments and further moving to real-time controlled dynamic environments.
Author Contributions:
Conceptualization, J.N.Y., M.-H.H., and M.M.Y.; methodology, J.N.Y., H.M.,
M.-H.H., and M.M.Y.; software, J.N.Y. and H.M.; validation, J.N.Y. and H.M.; writing-original draft
preparation, J.N.Y., M.-H.H., M.M.Y., and J.P.; writing-review and editing, M.-H.H., M.M.Y., and J.P.;
supervision, M.-H.H., M.M.Y. and J.P.; funding acquisition, M.-H.H. and J.P. All authors have read
and agreed to the published version of the manuscript.
Funding:
This work has been supported in part by the Academy of Finland-funded research project
314048 and Nokia Foundation (Award No. 20200147).
Data Availability Statement: Not applicable.
Conflicts of Interest: The authors declare no conflict of interest.
References
1.
Campion, M.; Ranganathan, P.; Faruque, S. A Review and Future Directions of UAV Swarm Communication Architectures.
In Proceedings of the 2018 IEEE International Conference on Electro/Information Technology (EIT), Rochester, MI, USA,
3–5 May 2018; pp. 903–908.
2.
Shakhatreh, H.; Sawalmeh, A.H.; Al-Fuqaha, A.; Dou, Z.; Almaita, E.; Khalil, I.; Othman, N.S.; Khreishah, A.; Guizani,
M. Unmanned Aerial Vehicles (UAVs): A Survey on Civil Applications and Key Research Challenges. IEEE Access
2019
,
7, 48572–48634. [CrossRef]
3.
Murray, R. Recent Research in Cooperative Control of Multi-Vehicle Systems. J. Dyn. Syst. Meas. Control
2007
,129, 571–598.
[CrossRef]
4.
He, L.; Bai, P.; Liang, X.; Zhang, J.; Wang, W. Feedback formation control of UAV swarm with multiple implicit leaders. Aerosp.
Sci. Technol. 2018,72, 327–334. [CrossRef]
5.
Besada, J.A.; Bergesio, L.; Campaña, I.; Vaquero-Melchor, D.; López-Araquistain, J.; Bernardos, A.M.; Casar, J.R. Drone Mission
Definition and Implementation for Automated Infrastructure Inspection Using Airborne Sensors. Sensors
2018
,18. [CrossRef]
[PubMed]
Remote Sens. 2021,13, 1059 16 of 17
6.
Mualla, Y.; Najjar, A.; Daoud, A.; Galland, S.; Nicolle, C.; Yasar, A.U.H.; Shakshuki, E. Agent-based simulation of unmanned
aerial vehicles in civilian applications: A systematic literature review and research directions. Future Gener. Comput. Syst.
2019
,
100, 344–364. [CrossRef]
7.
Gkiokas, A.; Cristea, A.I. Cognitive agents and machine learning by example: Representation with conceptual graphs. Comput.
Intell. 2018,34, 603–634. [CrossRef]
8. Dorri, A.; Kanhere, S.S.; Jurdak, R. Multi-Agent Systems: A Survey. IEEE Access 2018,6, 28573–28593. [CrossRef]
9.
Yasin, J.N.; Haghbayan, M.H.; Heikkonen, J.; Tenhunen, H.; Plosila, J. Formation Maintenance and Collision Avoidance in a
Swarm of Drones. In Proceedings of the 2019 3rd International Symposium on Computer Science and Intelligent Control, ISCSIC
2019, Amsterdam, The Netherlands, 25–27 September 2019; Association for Computing Machinery: New York, NY, USA, 2019.
[CrossRef]
10.
Zhuge, C.; Cai, Y.; Tang, Z. A Novel Dynamic Obstacle Avoidance Algorithm Based on Collision Time Histogram. Chin. J.
Electron. 2017,26, 522–529. [CrossRef]
11.
Wang, X.; Yadav, V.; Balakrishnan, S.N. Cooperative UAV Formation Flying With Obstacle/Collision Avoidance. IEEE Trans.
Control Syst. Technol. 2007,15, 672–679. [CrossRef]
12.
Choi, J.; Kim, Y. Fuel-Efficient Formation Flight-Control Design Based on Energy Maneuverability. J. Guid. Control Dyn.
2008
,
31, 1145–1150.
13.
Lin, Y.; Saripalli, S. Collision avoidance for UAVs using reachable sets. In Proceedings of the 2015 International Conference on
Unmanned Aircraft Systems, ICUAS 2015, Denver, CO, USA, 9–12 June 2015; Institute of Electrical and Electronics Engineers Inc.:
Piscataway, NJ, USA, 2015; pp. 226–235. [CrossRef]
14.
Esfahlani, S.S. Mixed reality and remote sensing application of unmanned aerial vehicle in fire and smoke detection. J. Ind. Inf.
Integr. 2019. [CrossRef]
15. Valavanis, K.P. Unmanned Aircraft Systems: The Current State-Of-The-Art; Springer: Berlin/Heidelberg, Germany, 2016.
16.
Wargo, C.A.; Church, G.C.; Glaneueski, J.; Strout, M. Unmanned Aircraft Systems (UAS) research and future analysis.
In Proceedings of the 2014 IEEE Aerospace Conference, Big Sky, MT, USA, 1–8 March 2014; pp. 1–16.
17.
Yasin, J.N.; Mohamed, S.A.S.; Haghbayan, M.H.; Heikkonen, J.; Tenhunen, H.; Plosila, J. Navigation of Autonomous Swarm of
Drones Using Translational Coordinates. In Advances in Practical Applications of Agents, Multi-Agent Systems, and Trustworthiness;
The PAAMS Collection; Demazeau, Y., Holvoet, T., Corchado, J.M., Costantini, S., Eds.; Springer International Publishing:
Cham, Switzerland, 2020; pp. 353–362.
18.
Madridano, A.; Al-Kaff, A.; Martín, D.; de la Escalera, A. 3D Trajectory Planning Method for UAVs Swarm in Building
Emergencies. Sensors 2020,20, 642. [CrossRef]
19.
Yasin, J.N.; Mohamed, S.A.S.; Haghbayan, M.; Heikkonen, J.; Tenhunen, H.; Plosila, J. Unmanned Aerial Vehicles (UAVs):
Collision Avoidance Systems and Approaches. IEEE Access 2020,8, 105139–105155. [CrossRef]
20.
Prats, X.; Delgado, L.; Ramirez, J.; Royo, P.; Pastor, E. Requirements, issues, and challenges for sense and avoid in unmanned
aircraft systems. J. Aircr. 2012,49, 677–687. [CrossRef]
21.
Ferrera, E.; Alcántara, A.; Capitán, J.; Castaño, A.; Marrón, P.; Ollero, A. Decentralized 3D Collision Avoidance for Multiple UAVs
in Outdoor Environments. Sensors 2018,18, 4101. [CrossRef]
22.
Yasin, J.N.; Mohamed, S.A.S.; Haghbayan, M.-H.; Heikkonen, J.; Tenhunen, H.; Yasin, M.M.; Plosila, J. Night vision obstacle detec-
tion and avoidance based on Bio-Inspired Vision Sensors. In Proceedings of the 2020 IEEE Sensors,
Rotterdam, The Netherlands,
25–28 October 2020; pp. 1–4. [CrossRef]
23.
Choi, D.; Lee, K.; Kim, D. Enhanced Potential Field-Based Collision Avoidance for Unmanned Aerial Vehicles in a Dynamic
Environment. In Proceedings of the AIAA Scitech 2020 Forum, Orlando, FL, USA, 6–10 January 2020. [CrossRef]
24.
Radmanesh, M.; Kumar, M.; Guentert, P.H.; Sarim, M. Overview of Path-Planning and Obstacle Avoidance Algorithms for UAVs:
A Comparative Study. Unmanned Syst. 2018,6, 95–118. [CrossRef]
25.
Seo, J.; Kim, Y.; Kim, S.; Tsourdos, A. Collision Avoidance Strategies for Unmanned Aerial Vehicles in Formation Flight. IEEE
Trans. Aerosp. Electron. Syst. 2017,53, 2718–2734. [CrossRef]
26.
Zhu, X.; Liang, Y.; Yan, M. A Flexible Collision Avoidance Strategy for the Formation of Multiple Unmanned Aerial Vehicles.
IEEE Access 2019,7, 140743–140754. [CrossRef]
27. Zhang, X.; Liniger, A.; Borrelli, F. Optimization-based collision avoidance. arXiv 2017, arXiv:1711.03449.
28.
Pham, H.; Smolka, S.A.; Stoller, S.D.; Phan, D.; Yang, J. A survey on unmanned aerial vehicle collision avoidance systems. arXiv
2015, arXiv:1508.07723.
29.
Smith, N.E.; Cobb, R.; Pierce, S.J.; Raska, V. Optimal collision avoidance trajectories via direct orthogonal collocation for
unmanned/remotely piloted aircraft sense and avoid operations. In Proceedings of the AIAA Guidance, Navigation, and Control
Conference, National Harbor, MD, USA, 13–17 January 2014; p. 966.
30.
Yasin, J.N.; Mohamed, S.A.S.; Haghbayan, M.H.; Heikkonen, J.; Tenhunen, H.; Yasin, M.M.; Plosila, J. Energy-Efficient Formation
Morphing for Collision Avoidance in a Swarm of Drones. IEEE Access 2020,8, 170681–170695. [CrossRef]
31.
Anderson, B.D.O.; Fidan, B.; Yu, C.; Walle, D. UAV Formation Control: Theory and Application. In Recent Advances in Learning
and Control; Blondel, V.D., Boyd, S.P., Kimura, H., Eds.; Springer: London, UK, 2008; pp. 15–33.
Remote Sens. 2021,13, 1059 17 of 17
32.
Hoang, V.T.; Phung, M.D.; Dinh, T.H.; Ha, Q.P. Angle-Encoded Swarm Optimization for UAV Formation Path Planning.
In Proceedings of the
2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Madrid, Spain,
1–5 October 2018; pp. 5239–5244.
33.
Wu, X.; Wang, S.; Xing, M. Observer-Based Leader-Following Formation Control for Multi-Robot With Obstacle Avoidance. IEEE
Access 2019,7, 14791–14798. [CrossRef]
34.
Ren, W. Consensus based formation control strategies for multi-vehicle systems. In Proceedings of the 2006 American Control
Conference, Minneapolis, MN, USA, 14–16 June 2006; p. 6.
35.
Low, C.B.; Ng, Q.S. A flexible virtual structure formation keeping control for fixed-wing UAVs. In Proceedings of the 2011 9th
IEEE International Conference on Control and Automation (ICCA), Santiago, Chile , 19–21 December 2011; pp. 621–626.
36. Oh, K.K.; Park, M.C.; Ahn, H.S. A survey of multi-agent formation control. Automatica 2015,53, 424–440. [CrossRef]
37.
Shen, D.; Sun, Z.; Sun, W. Leader-follower formation control without leader’s velocity information. Sci. China Inf. Sci.
2014
,
57, 1–12. [CrossRef]
38.
Han, Q.; Li, T.; Sun, S.; Villarrubia, G.; de la Prieta, F. “1-N” Leader-Follower Formation Control of Multiple Agents Based on
Bearing-Only Observation. In Advances in Practical Applications of Agents, Multi-Agent Systems, and Sustainability: The PAAMS
Collection; Demazeau, Y., Decker, K.S., Bajo Pérez, J., de la Prieta, F., Eds.; Springer International Publishing: Cham, Switzerland,
2015; pp. 120–130.
39.
Yasin, J.N.; Haghbayan, M.H.; Yasin, M.M.; Plosila, J. Swarm Formation Morphing for Congestion Aware Collision Avoidance.
arXiv 2020, arXiv:cs.RO/2011.03883.
40.
Bayezit, I.; Fidan, B. Distributed Cohesive Motion Control of Flight Vehicle Formations. IEEE Trans. Ind. Electron.
2013
,
60, 5763–5772. [CrossRef]
41.
Beard, R.W.; Lawton, J.; Hadaegh, F.Y. A coordination architecture for spacecraft formation control. IEEE Trans. Control Syst.
Technol.2001,9, 777–790. [CrossRef]
42.
Li, N.H.; Liu, H.H. Formation UAV flight control using virtual structure and motion synchronization. In Proceedings of the 2008
American Control Conference, IEEE, Seattle, WA, USA, 11–13 June 2008; pp. 1782–1787.
43.
Dong, L.; Chen, Y.; Qu, X. Formation Control Strategy for Nonholonomic Intelligent Vehicles Based on Virtual Structure and
Consensus Approach. Green Intelligent Transportation System and Safety. Procedia Eng. 2016,137, 415–424. [CrossRef]
44.
Lawton, J.R.; Beard, R.W.; Young, B.J. A decentralized approach to formation maneuvers. IEEE Trans. Robot. Autom.
2003
,
19, 933–941. [CrossRef]
45.
Balch, T.; Arkin, R.C. Behavior-based formation control for multirobot teams. IEEE Trans. Robot. Autom.
1998
,14, 926–939.
[CrossRef]
46.
Bartashevich, P.; Koerte, D.; Mostaghim, S. Energy-saving decision making for aerial swarms: PSO-based navigation in
vector fields. In Proceedings of the 2017 IEEE Symposium Series on Computational Intelligence (SSCI), Honolulu, HI, USA,
27 November–1 December 2017; pp. 1–8. [CrossRef]
47.
Majd, A.; Loni, M.; Sahebi, G.; Daneshtalab, M. Improving Motion Safety and Efficiency of Intelligent Autonomous Swarm of
Drones. Drones 2020,4, 48. [CrossRef]
48.
Tseng, C.M.; Chau, C.K.; Elbassioni, K.M.; Khonji, M. Flight tour planning with recharging optimization for battery-operated
autonomous drones. arXiv 2017, arXiv:1703.10049.
49.
Zorbas, D.; Razafindralambo, T.; Luigi, D.P.P.; Guerriero, F. Energy Efficient Mobile Target Tracking Using Flying Drones. Procedia
Comput. Sci. 2013,19, 80–87. [CrossRef]
50.
Al-Sabban, W.H.; Gonzalez, L.F.; Smith, R.N. Wind-energy based path planning for Unmanned Aerial Vehicles using Markov
Decision Processes. In Proceedings of the 2013 IEEE International Conference on Robotics and Automation, Karlsruhe, Germany,
6–10 May 2013; pp. 784–789. [CrossRef]
51.
Puck LITE Datasheets-Velodynelidar-PDF Catalogs-Technical Documentation. Available online: https://pdf.directindustry.com/
pdf/velodynelidar/puck-lite-datasheets/182407-676096.html (accessed on 16 February 2021)
52.
Legacy XBee S1 802.15.4 Product Datasheet. Available online: https://www.digi.com/resources/library/data-sheets/ds_
xbeemultipointmodules (accessed on 16 February 2021)
53.
Zhu, A.Z.; Thakur, D.; Özaslan, T.; Pfrommer, B.; Kumar, V.; Daniilidis, K. The Multivehicle Stereo Event Camera Dataset: An
Event Camera Dataset for 3D Perception. IEEE Robot. Autom. Lett. 2018,3, 2032–2039. [CrossRef]
... Minimization of the energy consumption of a swarm as a whole is another important research area with core emphasis on a varied set of themes, such as dealing with external influences [23,24], optimization of consumption due to ranging sensors [25], efficient interrobot communication [26], optimization of distance to be traveled [27,28], or recharging optimization [29,30]. In this respect, we present DCP-SLAM, a LiDAR-based distributed collaborative partial SLAM framework for swarm robotics. ...
... Navigation in environments with no prior map information raises several research challenges. The autonomous robots or agents, while navigating in unknown environments and not having to bank on acquiring information from central or remote servers, must utilize their onboard sensors to observe, analyze, and perform necessary actions based on the information at hand for collision-free navigation and successful completion of the mission [25]. At the same time, reducing the power consumed by the autonomous agents, i.e., effectively exploiting the available resources, in order to increase the mission duration is of paramount importance [32]. ...
Article
Full-text available
Collaborative robots represent an evolution in the field of swarm robotics that is pervasive in modern industrial undertakings from manufacturing to exploration. Though there has been much work on path planning for autonomous robots employing floor plans, energy-efficient navigation of autonomous robots in unknown environments is gaining traction. This work presents a novel methodology of low-overhead collaborative sensing, run-time mapping and localization, and navigation for robot swarms. The aim is to optimize energy consumption for the swarm as a whole rather than individual robots. An energy- and information-aware management algorithm is proposed to optimize the time and energy required for a swarm of autonomous robots to move from a launch area to the predefined destination. This is achieved by modifying the classical Partial Swarm SLAM technique, whereby sections of objects discovered by different members of the swarm are stitched together and broadcast to members of the swarm. Thus, a follower can find the shortest path to the destination while avoiding even far away obstacles in an efficient manner. The proposed algorithm reduces the energy consumption of the swarm as a whole due to the fact that the leading robots sense and discover respective optimal paths and share their discoveries with the followers. The simulation results show that the robots effectively re-optimized the previous solution while sharing necessary information within the swarm. Furthermore, the efficiency of the proposed scheme is shown via comparative results, i.e., reducing traveling distance by 13% for individual robots and up to 11% for the swarm as a whole in the performed experiments.
... where is the operational flight range, is the area coverage, is the camera focal length, is the UAV (sensor) flight altitude, is the camera CCD pixel size, ℎ is the number of pixels perpendicular to the flight direction, and is the side overlap. The formula derivation process is not detailed here, but can be found in the literature [34]. ...
... where l is the operational flight range, s is the area coverage, f is the camera focal length, H is the UAV (sensor) flight altitude, u is the camera CCD pixel size, h is the number of pixels perpendicular to the flight direction, and P h is the side overlap. The formula derivation process is not detailed here, but can be found in the literature [34]. ...
Article
Full-text available
Recently, unmanned aerial vehicle (UAV) remote sensing has been widely used in emergency scenarios; the operating mode has transitioned from one UAV to multiple UAVs. However, the current multiple-UAV remote sensing mode is characterized by high labor costs and limited operational capabilities; meanwhile, there is no suitable UAV swarm scheduling method that can be applied to remote sensing in emergency scenarios. To solve these problems, this study proposes a UAV swarm scheduling method. Firstly, the tasks were formulated and decomposed according to the data requirements and the maximum flight range of a UAV; then, the task sets were decomposed according to the maximum flight range of the UAV swarm to form task subsets; finally, aiming at the shortest total flight range of the task subsets and to balance the flight ranges of each UAV, taking the complete execution of the tasks as the constraint, the task allocation model was constructed, and the model was solved via a particle swarm optimization algorithm to obtain the UAV swarm scheduling scheme. Compared with the direct allocation method and the manual scheduling methods, the results show that the proposed method has high usability and efficiency.
... To maximize the sensing energy depending on the volume of environmental data gathered, an energyaware information management method is suggested to prevent over sensing. The self-awareness or self-localization properties of the swarm in the space domain dictate the information required from each unit [51]. ...
Article
Drone swarms, or unmanned aerial vehicles (UAVs), are becoming a promising domain in many areas of our lives. They are intricate, multidisciplinary systems, and most research projects concentrate on individual system components for particular use cases. Its involvement in missions and services has shown an imperatively positive influence. This review's objectives are to give a broad overview of the primary applications that spur most research efforts in this area. In our review, we have selected sixty articles between 2019 and 2024 about drone swarms. The review results outline the covered usage fields of drones in the chosen articles. It highlighted communication and control mostly in twenty-seven articles, services in nineteen articles, and tracking in fourteen articles, and we categorized the service domains more specifically as inventory, health care, defense, rescuing, and delivery. Besides that, the simulation aspect has been used for categorization as follows: twelve articles have specified their simulations, thirty-one articles haven’t specified their simulations, and seventeen articles haven’t used simulations. In addition, we have concluded the result of the SWOT analysis for the drone applications.
... Additionally, swarms can perform the so-called intelligence, surveillance, and reconnaissance (ISR) missions and multitarget operations at the same time, thanks to advanced cooperation approaches. So, the UAS swarm can be an adaptive system by fully detecting the surrounding inputs and communicating with the other systems [1,15]. ...
Article
The deployment of a drone swarm from carrier aircraft can support critical operations in which target aircraft should be protected by ground-based radar detection. To be successful as a countermeasure, the drones that are involved in the swarm must be equipped with proper payload systems and the geometry among the swarm units must be well designed. This paper describes the development of a scaled drone swarm geometry that can electromagnetically obscure a target from radar detection. Different drone swarm configurations were analyzed to select a feasible solution to develop ground and in-flight tests. During the tests, a compact software radar was used to detect a target drone. Several ground tests were executed to characterize the radar echo response of the target drone installed on a wooden tripod. Different geometries of corner reflectors were tested on tripods to select the swarm geometry that better obscured the target drone. Flight tests were executed involving decoy drones equipped with a corner reflector and a target drone to validate the proposed drone swarm geometry. The presented design of a drone swarm geometry allowed us to characterize the scaled distances among the swarm units and between the swarm and the target that must be protected against a ground-based radar detection.
... Additionally, swarms can perform the so-called Intelligence, Surveillance, and Reconnaissance (ISR) missions and multi target operations at the same time, thanks to advanced cooperation approaches. So, UAS swarm can be an adaptive system, by fully detecting the surrounding inputs and communicating with the other systems [1,15]. ...
Conference Paper
View Video Presentation: https://doi.org/10.2514/6.2022-0855.vid The paper describes the development of a drone swarm configuration that can electromagnetically obscure a target-drone from radar detection. At the beginning, different drone swarm architectures were analyzed in order to select the best solution to develop ground and in-flight tests. During the tests execution, a compact software radar was used to detect a target-drone. Moreover, trihedral corner reflectors were used to perform the electromagnetic obscuration of the target-drone. Several ground tests were executed to characterize the radar echo response of the target-drone installed on a wooden tripod. Then, the radar echo response of the target-drone was compared with the one obtained by the target-drone obscured by corner reflectors. Different configurations of corner reflectors were tested on wooden tripods to select the swarm configuration that better obscures the target-drone. At the end, flight tests were executed involving decoy-drones equipped with a corner reflector and a target-drone to assess and validate the developed drone swarm as countermeasure against ground-based radar detection.
Article
Full-text available
Collision avoidance is one of the most important topics in the robotics field. In this problem, the goal is to move the robots from initial locations to target locations such that they follow the shortest non-colliding paths in the shortest time and with the least amount of energy. Robot navigation among pedestrians is an example application of this problem which is the focus of this paper. This paper presents a distributed and real-time algorithm for solving collision avoidance problems in dense and complex 2D and 3D environments. This algorithm uses angular calculations to select the optimal direction for the movement of each robot and it has been shown that these separate calculations lead to a form of cooperative behavior among agents. We evaluated the proposed approach on various simulation and experimental scenarios and compared the results with ORCA one of the most important algorithms in this field. The results show that the proposed approach is at least 25% faster than ORCA while is also more reliable. The proposed method is shown to enable fully autonomous navigation of a swarm of Crazyflies.
Chapter
The focus of this work is to present a novel methodology utilizing the classical SLAM technique and integrating with the swarm agents for localizing, guiding, and retrieving the agents towards the optimal path while using only necessary tracker-based information between the agents. While navigating in an unknown environment with no-prior map information, upon encountering large obstacles (out of the field of view detection range of the onboard sensors, the swarm is divided into sub-swarms. This is done while dropping tracking points at every turn. Similarly, the time stamps for every turn taken and the gap width available between obstacles are recorded. Once an agent from any sub-swarm category reaches the destination, the agent broadcasts these tracker points to the rest of the swarm agents. Utilizing this broadcasted key information, the rest of the agents are able to navigate toward the destination without having to find the path. With the help of simulation examples, it is shown that the proposed technique is efficient over other similar randomized turn-based techniques.
Article
Full-text available
Unmanned aerial vehicles (UAVs) are widely applied in civil applications, such as disaster relief, agriculture and cargo transportation, and so on. With the massive number of UAV flight activities, the anti-collision technologies aiming to avoid the collisions between UAVs and other objects have attracted much attention. The anti-collision technologies are of vital importance to guarantee the survivability and safety of UAVs. In this article, a comprehensive survey on UAV anti-collision technologies is presented. We firstly introduce laws and regulations on UAV safety which prevent collision at the policy level. Then, the process of anti-collision technologies is reviewed from three aspects, i.e., obstacle sensing, collision prediction, and collision avoidance. We provide detailed survey and comparison of the methods of each aspect and analyze their pros and cons. Besides, the future trends on UAV anti-collision technologies are presented from the perspective of fast obstacle sensing and fast wireless networking. Finally, we summarize this article.
Article
Full-text available
The use of unmanned aerial vehicles (UAVs) is growing rapidly across many civil application domains, including real-time monitoring, providing wireless coverage, remote sensing, search and rescue, delivery of goods, security and surveillance, precision agriculture, and civil infrastructure inspection. Smart UAVs are the next big revolution in the UAV technology promising to provide new opportunities in different applications, especially in civil infrastructure in terms of reduced risks and lower cost. Civil infrastructure is expected to dominate more than $45 Billion market value of UAV usage. In this paper, we present UAV civil applications and their challenges. We also discuss the current research trends and provide future insights for potential UAV uses. Furthermore, we present the key challenges for UAV civil applications, including charging challenges, collision avoidance and swarming challenges, and networking and security-related challenges. Based on our review of the recent literature, we discuss open research challenges and draw high-level insights on how these challenges might be approached.
Article
Full-text available
The focus of this work is to present a novel methodology for optimal distribution of a swarm formation on either side of an obstacle, when evading the obstacle, to avoid overpopulation on the sides to reduce the agents' waiting delays, resulting in a reduced overall mission time and lower energy consumption. To handle this, the problem is divided into two main parts: 1) the disturbance phase: how to morph the formation optimally to avoid the obstacle in the least possible time in the situation at hand, and 2) the convergence phase: how to optimally resume the intended formation shape once the threat of potential collision has been eliminated. For the first problem, we develop a methodology which tests different formation morphing combinations and finds the optimal one, by utilizing trajectory, velocity, and coordinate information, to bypass the obstacle. For the second problem, we utilize a thin-plate splines (TPS) inspired temperature function minimization method to bring the agents back from the distorted formation into the desired formation in an optimal manner, after collision avoidance has been successfully performed. Experimental results show that, in the considered test scenario, the proposed approach results in substantial energy savings as compared with the traditional methods.
Article
Full-text available
Two important aspects in dealing with autonomous navigation of a swarm of drones are collision avoidance mechanism and formation control strategy; a possible competition between these two modes of operation may have negative implications for success and efficiency of the mission. This issue is exacerbated in the case of distributed formation control in leader-follower based swarms of drones since nodes concurrently decide and act through individual observation of neighbouring nodes’ states and actions. To dynamically handle this duality of control, a plan of action for multi-priority control is required. In this paper, we propose a method for formation-collision co-awareness by adapting the thin-plate splines algorithm to minimize deformation of the swarm's formation while avoiding obstacles. Furthermore, we use a non-rigid mapping function to reduce the lag caused by such maneuvers. Simulation results show that the proposed methodology maintains the desired formation very closely in the presence of obstacles, while the response time and overall energy efficiency of the swarm is significantly improved in comparison with the existing methods where collision avoidance and formation control are only loosely coupled. Another important result of using non-rigid mapping is that the slowing down effect of obstacles on the overall speed of the swarm is significantly reduced, making our approach especially suitable for time critical missions.
Article
Full-text available
Interest is growing in the use of autonomous swarms of drones in various mission-physical applications such as surveillance, intelligent monitoring, and rescue operations. Swarm systems should fulfill safety and efficiency constraints in order to guarantee dependable operations. To maximize motion safety, we should design the swarm system in such a way that drones do not collide with each other and/or other objects in the operating environment. On other hand, to ensure that the drones have sufficient resources to complete the required task reliably, we should also achieve efficiency while implementing the mission, by minimizing the travelling distance of the drones. In this paper, we propose a novel integrated approach that maximizes motion safety and efficiency while planning and controlling the operation of the swarm of drones. To achieve this goal, we propose a novel parallel evolutionary-based swarm mission planning algorithm. The evolutionary computing allows us to plan and optimize the routes of the drones at the run-time to maximize safety while minimizing travelling distance as the efficiency objective. In order to fulfill the defined constraints efficiently, our solution promotes a holistic approach that considers the whole design process from the definition of formal requirements through the software development. The results of benchmarking demonstrate that our approach improves the route efficiency by up to 10% route efficiency without any crashes in controlling swarms compared to state-of-the-art solutions.
Article
Full-text available
Moving towards autonomy, unmanned vehicles rely heavily on state-of-the-art collision avoidance systems (CAS). A lot of work is being done to make the CAS as safe and reliable as possible, necessitating a comparative study of the recent work in this important area. The paper provides a comprehensive review of collision avoidance strategies used for unmanned vehicles, with the main emphasis on unmanned aerial vehicles (UAV). It is an in-depth survey of different collision avoidance techniques that are categorically explained along with a comparative analysis of the considered approaches w.r.t. different scenarios and technical aspects. This also includes a discussion on the use of different types of sensors for collision avoidance in the context of UAVs.
Article
Full-text available
The development in Multi-Robot Systems (MRS) has become one of the most exploited fields of research in robotics in recent years. This is due to the robustness and versatility they present to effectively undertake a set of tasks autonomously. One of the essential elements for several vehicles, in this case, Unmanned Aerial Vehicles (UAVs), to perform tasks autonomously and cooperatively is trajectory planning, which is necessary to guarantee the safe and collision-free movement of the different vehicles. This document includes the planning of multiple trajectories for a swarm of UAVs based on 3D Probabilistic Roadmaps (PRM). This swarm is capable of reaching different locations of interest in different cases (labeled and unlabeled), supporting of an Emergency Response Team (ERT) in emergencies in urban environments. In addition, an architecture based on Robot Operating System (ROS) is presented to allow the simulation and integration of the methods developed in a UAV swarm. This architecture allows the communications with the MavLink protocol and control via the Pixhawk autopilot, for a quick and easy implementation in real UAVs. The proposed method was validated by experiments simulating building emergences. Finally, the obtained results show that methods based on probability roadmaps create effective solutions in terms of calculation time in the case of scalable systems in different situations along with their integration into a versatile framework such as ROS.
Conference Paper
Full-text available
Collision avoidance in aerial environments using the conventional artificial potential field(APF) often faces local minima problems and the results prevent unmanned aerial vehicles(UAVs) from performing their missions. In addition, an UAV’s paths planned based on theconventional method are safe trajectories only in a certain static environment. To generateoptimal and collision-free paths in a dynamic environment, the authors propose a novel APFapproach, called “enhanced curl-free vector field”. For the repulsive potential field of theapproach proposed, one computes each angle between the velocity vectors of UAVs and therelative position vectors of moving obstacles to the UAVs. The comparisons of the computedangles and the velocity of UAVs determine the direction of the curl-free vector field. Resultsfrom two case simulations, static obstacles with local minima and dynamic obstacles, show thatour approach solves the local minima problems of path planning and generates more efficientpaths for avoiding potential collisions caused by dynamic obstacles compared to the existingAPF methods.
Conference Paper
Moving towards autonomy, unmanned vehicles rely heavily on state-of-the-art collision avoidance systems (CAS). However, the detection of obstacles especially during night-time is still a challenging task since the lighting conditions are not sufficient for traditional cameras to function properly. Therefore, we exploit the powerful attributes of event-based cameras to perform obstacle detection in low lighting conditions. Event cameras trigger events asynchronously at high output temporal rate with high dynamic range of up to 120 dB. The algorithm filters background activity noise and extracts objects using robust Hough transform technique. The depth of each detected object is computed by triangulating 2D features extracted utilising LC-Harris. Finally, asynchronous adaptive collision avoidance (AACA) algorithm is applied for effective avoidance. Qualitative evaluation is compared using event-camera and traditional camera.
Book
This book constitutes the proceedings of the 18th International Conference on Practical Applications of Agents and Multi-Agent Systems, PAAMS 2020, held in L'Aquila, Italy, in October 2020. The 29 regular and 17 demo papers presented in this volume were carefully reviewed and selected from 64 submissions. They deal with the application and validation of agent-based models, methods, and technologies in a number of key applications areas, including: advanced models and learning, agent-based programming, decision-making, educa-tion and social interactions, formal and theoretic models, health and safety, mobility and the city, swarms and task allocation.
Conference Paper
This work focuses on an autonomous swarm of drones, a multi-agent system, where the leader agent has the capability of intelligent decision making while the other agents in the swarm follow the leader blindly. The proposed algorithm helps with cost cutting especially in the multi-drone systems, i.e., swarms, by reducing the power consumption and processing requirements of each individual agent. It is shown that by applying a pre-specified formation design with feedback cross-referencing between the agents, the swarm as a whole can not only maintain the desired formation and navigate but also avoid collisions with obstacles and other drones. Furthermore, the power consumed by the nodes in the considered test scenario, is reduced by 50% by utilising the proposed methodology.