Sweeping-Based Multi-Robot Exploration in an

Unknown Environment Using Webots

Nirali Sanghvi

1 a

, Rajdeep Niyogi

1 b

and Alfredo Milani

2 c

Department of Computer Science and Engineering, Indian Institute of Technology, Roorkee, India

Department of Computer Science and Mathematics, University of Perugia, Italy

Keywords:

Exploration, Unknown Environment, Multi-Robot System, Coverage.

Abstract:

In this paper, a sweeping algorithm is proposed with the goal of achieving maximal coverage while minimizing

the overlapping areas, in an unknown environment. Two scenarios are considered: one in which the robots

do not communicate with one another, and another in which the robots are allowed to communicate with

one another. The simulations are performed on Webots, a multi-robot simulator, varying various parameters

like environment size, obstacles, and number of robots and their positions. The coverage obtained with the

proposed approach is 89-98%. When the robots are allowed to communicate, there is a reduction in exploration

time that ranges from a minimum of 33% to a maximum of 68%.

1 INTRODUCTION

Multi-robot systems (MRS) have garnered signiﬁcant

attention from the research community and engineer-

ing practitioners due to their ability to enhance efﬁ-

ciency and reduce the human workload. These sys-

tems have proven highly valuable in various applica-

tions, such as target searching, structural inspection,

and boundary monitoring, where exploring unknown

environments is a critical challenge (de Almeida et al.,

2019), (Hayajneh and Al Mahasneh, 2022). The

primary objective of exploration is to guide a robot

through unfamiliar or uncharted areas without prior

knowledge or initial parameters. In many practical

scenarios, such as military missions, space explo-

ration, search and rescue efforts (Yanguas-Rojas and

Mojica-Nava, 2017), and agricultural work (Bechar

and Vigneault, 2017), the successful execution of

tasks often relies on effective environmental explo-

ration (Cao et al., 2023). Compared to single-

robot systems, multi-robot systems are frequently em-

ployed in these complex and hazardous contexts. This

preference arises from their notable attributes, includ-

ing robust adaptability, exceptional ﬂexibility, and a

high degree of reliability (Wang et al., 2016). In a

MRS, multiple robots collaborate to achieve a com-

https://orcid.org/0009-0003-4245-0587

https://orcid.org/0000-0003-1664-4882

https://orcid.org/0000-0003-4534-1805

mon goal while pursuing individual tasks within the

same environment (Sabattini et al., 2017).

The challenges in multi-robot coverage can gener-

ally be categorized into two domains: Coverage Path

Planning (CPP) (Galceran and Carreras, 2013) and

Coverage Control problems (Savkin et al., 2015). In

multi-robot CPP, the focus is on designing obstacle-

free paths that enable the accumulation of sensor foot-

prints from the robots to effectively cover a desig-

nated area or volume. Depending on the speciﬁc

CPP tasks, efﬁciency metrics may be deﬁned based on

factors such as coverage percentage (Doitsidis et al.,

2012), and time to completion (Avellar et al., 2015).

On the other hand, in coverage control problems, the

objective is to develop distributed control laws for the

robots that maximize certain coverage criteria, such

as coverage frequency. Initially, sweep coverage was

addressed as a coverage control problem, aiming to

optimize the detection rate of events during periodic

coverage missions within a region (Gage, 1992).

Different strategies have been developed to solve

the exploration problem in an unknown environment

(Sharma and Tiwari, 2016). The most basic method

of exploration is exploring random points in the en-

vironment, say, within some range of the robot. It is

based on randomness of the selection of the points.

Improvised versions of this method involve picking

up certain points. Another method is a frontier-based

method where the boundary between the known and

unknown areas of the environment is explored and

248

Sanghvi, N., Niyogi, R. and Milani, A.

Sweeping-Based Multi-Robot Exploration in an Unknown Environment Using Webots.

DOI: 10.5220/0012343400003636

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 16th International Conference on Agents and Artiﬁcial Intelligence (ICAART 2024) - Volume 1, pages 248-255

ISBN: 978-989-758-680-4; ISSN: 2184-433X

eventually, the process continues till the entire area

is explored (Sharma and Tiwari, 2016). Another ap-

proach is a human-directed approach wherein humans

can direct the robots based on the information gath-

ered by a graphical user (Sharma and Tiwari, 2016).

One of the strategies is to divide the area into smaller

regions like using Voronoi partitions and then dynam-

ically assign the robots to explore those areas (Hu

et al., 2020). Similarly, sweeping is one of the strate-

gies that is used to achieve exploration and coverage

in an unknown environment.

In this paper, we present a sweeping-based al-

gorithm for maximizing coverage while minimizing

overlapping. We consider two scenarios: one in

which the robots do not communicate with one an-

other, and one in which the robots are allowed to com-

municate with one another. Extensive simulations are

used to examine the impact of communication. A de-

tailed discussion of the similarities and differences

between our approach and some other existing ap-

proaches is given in Section 2.

The rest of this article is structured as follows. Re-

lated work is discussed in Section 2. Section 3 pro-

vides an overview of the simulated environment and

the mobile robots used. The proposed algorithm is

given in Section 4. Simulation results are given in

Section 5. Conclusions are given in Section 6.

2 RELATED WORK

Over the past few decades, numerous researchers

have been interested in the exploration tasks of multi-

robot systems. A signiﬁcant portion of this work

builds upon the concept of “frontier”, initially pro-

posed in (Yamauchi, 1998). In this context, a frontier

is deﬁned as the boundary that separates unexplored

and explored accessible areas within an unknown en-

vironment, typically represented using an occupancy

grid map. Yamauchi’s pioneering work led to the de-

velopment of a well-known multi-robot exploration

approach, building upon his prior research. This ap-

proach, while effective in its own right, relies on a

somewhat greedy strategy and lacks robust collabo-

ration mechanisms. Consequently, there is a poten-

tial for robots to end up exploring the same frontiers

within the environment inadvertently (Li et al., 2019).

Various strategies have been devised to explore

unknown environments effectively. One approach

involves dividing the area into smaller partitions,

while another popular method employs waypoints

that guide the robot through the entire area (Ka-

malova et al., 2020). Additionally, biologically in-

spired algorithms are presented in (Kamalova et al.,

2020), (de Almeida et al., 2019) that utilize way-

points for exploration. This approach leverages

swarm-based strategies, allowing agents to navigate

efﬁciently to areas requiring coverage (Atınc¸ et al.,

2020). To handle uncertainty resulting from random

workload distribution, a decentralized workload par-

tition algorithm was introduced in (Zhai and Hong,

2013). This innovative approach entails segmenting

the target region into distinct stripes and ensuring

an equitable distribution of workload across each of

these stripes.

In recent years, some works have approached the

multi-robot sweep coverage problem as a one-time

coverage task, resembling CPP problems, with the

goal of maximizing coverage percentage (Shi et al.,

2018) or minimizing the time required for operation

(Zhai, 2014). Multi-robot sweep coverage is the task

of moving a group of robots to fully cover a des-

ignated region or space (Savkin et al., 2015). In a

broader context, the robots are granted the freedom to

move autonomously, either for one-time or periodic

coverage of a region. Their primary objective is to op-

timize a performance metric, such as coverage rate or

mission duration (Kong et al., 2006), (Senthilkumar

and Bharadwaj, 2012), (Rosalie et al., 2017), (Huang

et al., 2019).

In (Tran et al., 2022), the researchers have imple-

mented a sweeping algorithm for exploration. This

approach is characterized by a swarm-based strategy,

where a group of robots collaboratively explores fron-

tiers within the environment. As one set of fron-

tiers is successfully explored, the swarm moves on

to uncover the next set of frontiers and uses ROS

for the implementation. On a related note, (Zhang

and Noguchi, 2017) developed a multi-robot trac-

tor designed for agricultural tasks, utilizing the same

sweeping algorithm for exploration. Notably, their

approach assumes that all robots have a similar orien-

tation, a condition that aligns with one of the scenar-

ios considered in our algorithm, and they have used

a simulation software named Multi checker, which

is a Windows console application. In (Cao et al.,

2023), the sweeping algorithm is extended to involve

the dynamic division of the environment into distinct

stripes. These stripes are subsequently explored by

robots, thus exploring the environment, using ROS

and MATLAB for their simulations. Orientation same

means all the robots are having a same alignment and

the are also having similar motion. In (Zhang and

Noguchi, 2017), they have used this approach where

the orientation is same. In this paper, the orientation

is different for the robots.

Table 1 provides an overview of the essential char-

acteristics of the referenced studies, drawing com-

Sweeping-Based Multi-Robot Exploration in an Unknown Environment Using Webots

249

Table 1: Comparison of our work with existing works.

Main Characteristics A B C D

Sweeping method Frontier-based Partition-based Orientation same Orientation different

Methodology Swarm MRS MRS MRS

Number of robots 40-80 4-6 3-7 2-5

Communication Implicit Explicit Explicit Explicit

Coverage % 100 - 83-89 89-98

Tool ROS ROS Multi-checker Webots

A: (Tran et al., 2022) B: (Cao et al., 2023) C: (Zhang and Noguchi, 2017) D: In this work

parisons to the present research. It underscores both

commonalities and points of deviation or complemen-

tary features.

• All these works, including our work, adopt a de-

centralized approach.

• All these works, including our work, consider a

continuous environment.

• Our methodology is based on a Multi-Robot Sys-

tem (MRS), whereas (Tran et al., 2022) relies on

a swarm-based approach. In MRS, the number of

robots is limited whereas swarm robotics typically

consists of a large number of robots (Farinelli

et al., 2004). MRS-based approaches use explicit

communication (via the exchange of messages),

whereas swarm robotics use implicit communica-

tion (e.g., via pheromones). Unlike the previously

mentioned works, the existing literature does not

extensively address the potential implications of

communication in the context of exploring un-

known environments.

• Coverage % is indicated as in the papers, and in

(Cao et al., 2023) coverage is not mentioned, it is

shown as ’-’.

• (Tran et al., 2022) implements a frontier-based

sweeping approach. (Cao et al., 2023) employs

a partition-based sweeping approach. Our al-

gorithm involves a general sweeping-based ap-

proach. In (Tran et al., 2022), the number of

robots is dependent on the number of turns of the

sweep-based algorithm, but for our approach, it is

independent of any such parameters.

3 ENVIRONMENT DESCRIPTION

We consider a closed environment that comprises

multiple robots and obstacles of different shapes and

sizes. The environment is represented as a continuous

environment. The robots used are differential drive

robots. The robots interact with the environment and

acquire data through their sensors. This data is uti-

lized for tasks such as collecting samples, images,

etc., all of which ultimately contribute to the process

of exploration. The robots can communicate among

themselves as shown in Figure 6. Robots aim to ex-

plore the environment and minimize the overlapping

of the areas explored by them.

In this work, the robots are considered to be ho-

mogeneous. Obstacles in the environment fall into

two categories: big obstacles and small obstacles. Big

obstacles are deﬁned as those whose dimensions ex-

ceed twice the size of the robot while any other obsta-

cles are categorized as small obstacles; a robot cannot

explore the area occupied by a big obstacle. While

exploring, a robot considers another robot as an ob-

stacle.

This paper uses the open-source robot simulator

Webots (Michel, 2004) for conducting simulations.

Webots is one of the versatile open-source simula-

tors that are available for academic purposes (Ramli

et al., 2015). Several researchers have used Webots

for their research like (Stan and Oprea, 2019), (Han

et al., 2019), (Rangu et al., 2023), etc. For our ex-

periments, we have employed E-puck robots, which

are supported in Webots (Figure 1 (Mondada et al.,

2009)).

Figure 1: E-puck robot (Mondada et al., 2009).

ICAART 2024 - 16th International Conference on Agents and Artiﬁcial Intelligence

250

4 PROPOSED APPROACH

In this section, we present a sweeping-based algo-

rithm, that takes as inputs the initial and goal positions

of a robot.

Algorithm 1: Sweeping Based Algorithm.

Data: initial-position, ﬁnal-position

Result: Environment is explored

1 current-position := initial-position;

2 while current-position ̸= ﬁnal-position do

3 s := SWEEP(current-position);

4 current-position := ORIENT(s);

5 SWEEP(position curr) {

6 EXPLORE(curr);

7 curr

′

:= generate-successor(curr);

8 if curr

′

0 then

9 return curr and exit;

10 else

11 SWEEP(curr

′

) }

The sweep-based algorithm operates as follows:

Initially, we designate the robot’s initial position as

the current position. If the current position does not

match the ﬁnal position, we invoke the SWEEP func-

tion. The SWEEP function calls the EXPLORE func-

tion that does essential tasks that extends beyond mere

traversal, such as collecting samples or performing

speciﬁc actions, as indicated by the context (Nesnas

et al., 2021) at the current position. Once the cur-

rent position has been thoroughly explored, the algo-

rithm proceeds to generate the successor of the cur-

rent position. Generate-successor returns the succes-

sor if one exists, otherwise, it returns an empty set.

SWEEP explores recursively until a new position can-

not be found.

The ORIENT function is responsible for aligning

the robot in the correct direction for the exploration

to continue and then moving it at a distance equal to

twice its sensor range. This step ensures comprehen-

sive coverage of the entire region, leaving no unex-

plored areas in between. Once the ORIENT function

concludes, it returns the point at which the orienta-

tion process ends which is assigned to the current po-

sition. This point is subsequently passed as input to

the SWEEP function, initiating another iteration of the

exploration process. The algorithm continues till it

reaches the goal position and then terminates.

In Figures 2, 3, and 4 we present a detailed step-

by-step illustration of the algorithm’s functioning. Let

us consider a 5 × 5 grid, where each cell represents a

speciﬁc location, for the illustration of the algorithm.

The initial position and goal position as shown in

Figure 2 are given as input to the algorithm. Now,

the algorithm starts. First of all, it assigns the initial

position to the current position and checks whether it

is equal to the ﬁnal position.

Figure 2: Beginning of the sweeping-based algorithm. Ini-

tial location I, Goal location G.

Figure 3: ORIENT function executed. Gray denotes ex-

plored area, Yellow denotes orientation.

Now, if it is not equal to the ﬁnal position, it en-

ters the while loop. SWEEP function is called with

the current position as the argument. Now, as the

SWEEP function starts, it calls the EXPLORE func-

tion at the current position. EXPLORE function com-

pletes the tasks like collecting samples, collecting

data, etc whatever is assigned to it. Now the generate-

successor function is called to generate the successor

of the current position. As shown in Figure 3, the area

where exploration has been completed is shaded with

gray color. The area where exploration has not yet

been done is shaded in white color.

The SWEEP function continues till generate-

successor does not return an empty set. Then at the

end of the column when generate-successor function

is called, it returns an empty set as there is a boundary

and no further area to explore at that point. So, the

SWEEP function terminates.

Now, the output of the SWEEP function is passed

Sweeping-Based Multi-Robot Exploration in an Unknown Environment Using Webots

251

Figure 4: Goal reached and the algorithm terminates.

as an argument to the ORIENT function. ORIENT

changes the orientation of the robot to the point

where we can continue our motion. So, as shown in

Figure 3, the yellow color shows the position where

the ORIENT function has ended and the position is

returned. Now, the position returned is assigned to

the current position and the while loop continues.

Figure 4 shows the end result of our algorithm.

Case 1: The environment has no obstacles. The

algorithm ensures full coverage.

Case 2: The environment has only small obsta-

cles. The algorithm ensures full coverage.

Property of the Algorithm: If the environment

has only small obstacles or no obstacles, the algorithm

ensures complete coverage.

5 SIMULATION RESULTS

In this section, we present the simulation results il-

lustrating the performance of our algorithm. To con-

duct these experiments, we utilized the R2023b ver-

sion of the Webots simulator. The simulations have

been performed on a system with 11

Gen Intel(R)

Core(TM) i5-1155G7 processor with 16 GB RAM,

2.5GHz CPU, and 64-bit Windows operating system.

In our simulations, we have explored various scenar-

ios and collected corresponding results. These sim-

ulations encompass a range of parameters, including

environment size, obstacles, the number of robots de-

ployed, and the initial and destination points of the

robots.

For a ﬁxed conﬁguration of the obstacles, the

number of robots is varied. Each conﬁguration of

the obstacles is obtained by randomly placing some

random number of obstacles. The simulations are re-

peated 20 times. Each row of the Tables, given below,

shows the average of the values obtained from the 20

simulations. The environment size is taken as k × k,

where k = 1,2,5, 10 meter. The number of robots de-

ployed n is taken as: n = 2,3,4,5.

Meaning of the Different Parameters:

1. Time: Let t

be the time taken by robot i to reach

its goal position from the initial position. T is de-

ﬁned as the maximum time taken by any robot,

i.e., T = max{t

,.. .,t

}

In the Tables given below, t woc and t wc repre-

sent T without communication and with commu-

nication respectively.

2. Coverage: Let C

be the area covered by robot i.

Coverage, denoted by Cg, is the union of the areas

covered by each robot, i.e., Cg =

i=1

3. Overlap: It is the same area that is explored by

more than one robot, i.e., the function EXPLORE

is invoked by multiple robots. Overlap is denoted

by Op.

5.1 Without communication

In our ﬁrst simulation, we consider the scenario where

the robots do not communicate with each other. The

E-puck robot is equipped with eight infrared proxim-

ity sensors strategically positioned around its body.

(Mondada et al., 2009). These sensors play a pivotal

role in gauging the proximity of obstacles within the

robot’s surroundings

Figure 5: Path of the robots without communication in a

1m × 1m environment.

As illustrated in Figure 5 the path of Robot 1 is

depicted in blue, while the path of Robot 2 is rep-

resented in green. Robot 1 follows a vertical path,

while Robot 2 adopts a horizontal path. The entire

environment is effectively covered by both robots,

demonstrating the comprehensive coverage achieved

through our algorithm. If two robots are just travers-

ing through the same area either to orient or to align

themselves, then it is not considered as overlap.

ICAART 2024 - 16th International Conference on Agents and Artiﬁcial Intelligence

252

The robot can dynamically adjust its path to nav-

igate around the obstacle, thereby avoiding collisions

and progressing toward its intended goal. This sensor-

driven adaptive behavior continues until the robot

successfully reaches its predeﬁned destination. The

robot’s turning radius is set at twice the range of its

sensors. This deliberate choice yields excellent cov-

erage results and allows us to effectively navigate

through the environment while avoiding unexplored

gaps between paths.

Figure 5 provides a visual representation of how

the robots interact with the environment. When a

robot encounters an obstacle along its path, it adjusts

its route to bypass the obstacle. The path of Robot 1 is

indicated by the blue line, while the path of Robot 2 is

represented by the green line. This dynamic adapta-

tion ensures that the robots can effectively maneuver

around obstacles while adhering to our algorithm’s

guidelines.

In this phase, we have escalated the complexity by

varying the size of the environment and also increas-

ing the number of robots accordingly. Furthermore,

we have introduced an element of unpredictability by

randomly placing obstacles of varying sizes within

these environments. This comprehensive evaluation

allows us to see how the presence of obstacles impacts

critical metrics, including coverage, overlapping, and

the execution time of the algorithm.

Tables 2, 3, 4, and 5 below are the outcomes de-

rived from these simulations, shedding light on the

algorithm’s performance under these changing condi-

tions.

Table 2: Performance for environment-size 1m ×1m.

n t woc(s) t wc(s) Rd % Cg % Op %

2 0.38 0.12 67.87 93.75 56.25

3 0.50 0.17 65.63 95.31 100

4 0.57 0.22 61.11 96.88 100

5 0.65 0.26 59.87 97.18 100

Table 3: Performance for environment size 2m ×2m.

n t woc(s) t wc(s) Rd % Cg % Op %

2 1.39 0.56 59.71 89.06 78.13

3 1.43 0.63 55.94 92.19 89.06

4 1.49 0.68 54.36 93.75 100

5 1.55 0.71 54.19 95.31 100

In the Tables, n represents the number of robots

used. It can be seen in this experiment where there is

no communication between the robots, there is a lot of

overlapping area which leads to the wastage of com-

putational power and also increases the time taken by

Table 4: Performance for environment size 5m ×5m.

n t woc(s) t wc(s) Rd % Cg % Op %

2 5.10 3.10 39.24 91.25 90.50

3 5.21 3.15 39.48 92.75 95.25

4 5.35 3.38 36.80 95.28 100

5 5.43 3.65 32.71 96.75 100

Table 5: Performance for environment size 10m ×10m.

n t woc(s) t wc(s) Rd % Cg % Op %

2 20.15 12.27 39.11 94.88 95.13

3 21.60 13.35 38.17 95.63 97.56

4 22.86 14.83 35.13 98.13 100

5 24.66 15.53 37.03 98.36 100

the robots to reach their goal positions. Also, it can be

observed that at certain places, the time taken by three

robots is more than the time taken by two robots. This

happens especially in small environments where the

increasing number of robots adds up to the amount of

collision, leading to an increase in the overall amount

of time.

As the number of robots n is increased, the time

taken for exploration, coverage, and overlap also in-

creases. As n is increased, one robot acts as an ob-

stacle for the other, and hence the time increases.

With more robots, more area would be covered, which

means that there would be more overlap.

5.2 With Communication

In these experiments, the robots can communicate

with each other. Figure 6 shows how communication

takes place between the robots. As shown in Figure 6,

as soon as Robot 1 enters the range of Robot 2, they

exchange their identiﬁers (represented by a unique

number) using the Beacon signals (Gerasenko et al.,

2001). Now, say, Robot 2 has a higher identiﬁer than

Robot 1, then Robot 2 will continue its exploration

for the next step and Robot 1 will change its orienta-

tion and continue as per the algorithm, given in Sec-

tion 4. The objective of these experiments is to assess

the impact of communication in this setting. Figure 7

illustrates the paths the robots will follow in this en-

vironment.

Figure 7 shows the path that will be followed by

Robot 1 and Robot 2 by using communication be-

tween them. It can be seen that there is a signiﬁcant

amount of reduction in the overlapping areas. There

is no region that is explored more than once. While

changing the orientation, an area may be traversed

by more than one robot but not explored by multiple

robots.

The Tables 2, 3, 4, and 5, given above, should be

Sweeping-Based Multi-Robot Exploration in an Unknown Environment Using Webots

253

Figure 6: Communication between robots.

Figure 7: Path of the robots with communication in a

1m × 1m environment.

read as follows. t wc is the time taken with communi-

cation. The coverage is the same as without commu-

nication. Now there is no overlap. For different envi-

ronment sizes, the minimum and maximum reduction

in time with respect to the time without communica-

tion is given in Table 6. The reduction is calculated as

((t woc −t wc)/t woc) × 100. From Table 6, we ﬁnd

that the reduction in exploration time ranges from a

maximum of 68% to a minimum of 33%.

Table 6: Minimum/Maximum time reduction with commu-

nication.

Env-size Rd min % Rd max %

1m × 1m 59.87 67.87

2m × 2m 54.19 59.71

5m × 5m 32.71 39.48

10m × 10m 35.13 39.11

6 CONCLUSIONS

In this paper, a sweeping-based approach is devel-

oped for exploring an unknown environment using

multiple robots. To validate the approach, a multi-

robot simulator, Webots, is used. Extensive simula-

tions were conducted with varying environment sizes,

obstacles, the number of robots deployed, and the ini-

tial and destination location of the robots. The results

demonstrate that the proposed algorithm performs as

expected. The coverage obtained is 89–98%. When

the robots are allowed to communicate, there is a sig-

niﬁcant reduction in exploration time that ranges from

a maximum of 68% to a minimum of 33%. As part of

future work, the scope of the approach in smart farm-

ing would be explored.

ACKNOWLEDGMENT

The authors thank the anonymous reviewers for their

valuable comments that were helpful for improving

the paper. The second author was in part supported

by a research grant from Google.

REFERENCES

Atınc¸, G. M., Stipanovi

c, D. M., and Voulgaris, P. G.

(2020). A swarm-based approach to dynamic cov-

erage control of multi-agent systems. Automatica,

112:108637.

Avellar, G. S., Pereira, G. A., Pimenta, L. C., and Is-

cold, P. (2015). Multi-uav routing for area coverage

and remote sensing with minimum time. Sensors,

15(11):27783–27803.

Bechar, A. and Vigneault, C. (2017). Agricultural robots

for ﬁeld operations. part 2: Operations and systems.

Biosystems engineering, 153:110–128.

Cao, M., Cao, K., Li, X., and Xie, L. (2023). Distributed

control of multirobot sweep coverage over a region

with unknown workload distribution. IEEE Trans-

actions on Systems, Man, and Cybernetics: Systems,

53(10):6503–6515.

de Almeida, J. P. L. S., Nakashima, R. T., Neves-Jr, F., and

de Arruda, L. V. R. (2019). Bio-inspired on-line path

planner for cooperative exploration of unknown envi-

ronment by a multi-robot system. Robotics and Au-

tonomous Systems, 112:32–48.

Doitsidis, L., Weiss, S., Renzaglia, A., Achtelik, M. W.,

Kosmatopoulos, E., Siegwart, R., and Scaramuzza, D.

(2012). Optimal surveillance coverage for teams of

micro aerial vehicles in gps-denied environments us-

ing onboard vision. Autonomous Robots, 33:173–188.

Farinelli, A., Iocchi, L., and Nardi, D. (2004). Multi-

robot systems: a classiﬁcation focused on coordina-

ICAART 2024 - 16th International Conference on Agents and Artiﬁcial Intelligence

254

tion. IEEE Transactions on Systems, Man, and Cy-

bernetics, Part B (Cybernetics), 34(5):2015–2028.

Gage, D. W. (1992). Command control for many-robot sys-

tems. Unmanned Systems Magazine, 10(4):28–34.

Galceran, E. and Carreras, M. (2013). A survey on coverage

path planning for robotics. Robotics and Autonomous

systems, 61(12):1258–1276.

Gerasenko, S., Joshi, A., Rayaprolu, S., Ponnavaikko, K.,

and Agrawal, D. P. (2001). Beacon signals: what,

why, how, and where? Computer, 34(10):108–110.

Han, Q., Sun, S., and Lang, H. (2019). Leader-follower

formation control of multi-robots based on bearing-

only observations. International Journal of Robotics

and Automation, 34(2).

Hayajneh, M. and Al Mahasneh, A. (2022). Guidance, nav-

igation and control system for multi-robot network

in monitoring and inspection operations. Drones,

6(11):332.

Hu, J., Niu, H., Carrasco, J., Lennox, B., and Arvin, F.

(2020). Voronoi-based multi-robot autonomous ex-

ploration in unknown environments via deep rein-

forcement learning. IEEE Transactions on Vehicular

Technology, 69(12):14413–14423.

Huang, L., Zhou, M., Hao, K., and Hou, E. (2019). A sur-

vey of multi-robot regular and adversarial patrolling.

IEEE/CAA Journal of Automatica Sinica, 6(4):894–

903.

Kamalova, A., Kim, K. D., and Lee, S. G. (2020). Way-

point mobile robot exploration based on biologically

inspired algorithms. IEEE Access, 8:190342–190355.

Kong, C. S., Peng, N. A., and Rekleitis, I. (2006). Dis-

tributed coverage with multi-robot system. In Pro-

ceedings 2006 IEEE International Conference on

Robotics and Automation, 2006. ICRA 2006., pages

2423–2429. IEEE.

Li, G., Zhang, D., and Shi, Y. (2019). An unknown envi-

ronment exploration strategy for swarm robotics based

on brain storm optimization algorithm. In 2019 IEEE

Congress on Evolutionary Computation (CEC), pages

1044–1051. IEEE.

Michel, O. (2004). Cyberbotics ltd. webots™: professional

mobile robot simulation. International Journal of Ad-

vanced Robotic Systems, 1(1):5.

Mondada, F., Bonani, M., Raemy, X., Pugh, J., Cianci, C.,

Klaptocz, A., Magnenat, S., Zufferey, J.-C., Floreano,

D., and Martinoli, A. (2009). The e-puck, a robot de-

signed for education in engineering. In Proceedings of

the 9th conference on autonomous robot systems and

competitions, volume 1, pages 59–65. IPCB: Instituto

Polit

ecnico de Castelo Branco.

Nesnas, I. A., Fesq, L. M., and Volpe, R. A. (2021). Auton-

omy for space robots: Past, present, and future. Cur-

rent Robotics Reports, 2(3):251–263.

Ramli, N. R., Razali, S., and Osman, M. (2015). An

overview of simulation software for non-experts to

perform multi-robot experiments. In 2015 Interna-

tional Symposium on Agents, Multi-Agent Systems

and Robotics (ISAMSR), pages 77–82.

Rangu, G., Kulkarni, D. D., Nair, J. S., and Nair, S. B.

(2023). A hybrid federated reinforcement learning ap-

proach for networked robots. In International Confer-

ence on Science, Technology and Engineering, pages

493–500. Springer.

Rosalie, M., Dentler, J. E., Danoy, G., Bouvry, P., Kan-

nan, S., Olivares-Mendez, M. A., and Voos, H. (2017).

Area exploration with a swarm of uavs combining de-

terministic chaotic ant colony mobility with position

mpc. In 2017 International Conference on Unmanned

Aircraft Systems (ICUAS), pages 1392–1397. IEEE.

Sabattini, L., Secchi, C., and Fantuzzi, C. (2017). Multi-

robot systems implementing complex behaviors under

time-varying topologies. European Journal of Con-

trol, 38:73–87.

Savkin, A. V., Cheng, T. M., Xi, Z., Javed, F., Matveev,

A. S., and Nguyen, H. (2015). Decentralized cover-

age control problems for mobile robotic sensor and

actuator networks. John Wiley & Sons.

Senthilkumar, K. and Bharadwaj, K. K. (2012). Multi-

robot exploration and terrain coverage in an unknown

environment. Robotics and Autonomous Systems,

60(1):123–132.

Sharma, S. and Tiwari, R. (2016). A survey on multi

robots area exploration techniques and algorithms.

In 2016 International Conference on Computational

Techniques in Information and Communication Tech-

nologies (ICCTICT), pages 151–158. IEEE.

Shi, M., Qin, K., and Liu, J. (2018). Cooperative multi-

agent sweep coverage control for unknown areas of

irregular shape. IET Control Theory & Applications,

12(14):1983–1994.

Stan, A.-C. and Oprea, M. (2019). A case study of multi-

robot systems coordination using pso simulated in we-

bots. In 2019 11th International Conference on Elec-

tronics, Computers and Artiﬁcial Intelligence (ECAI),

pages 1–5. IEEE.

Tran, V. P., Garratt, M. A., Kasmarik, K., Anavatti, S. G.,

and Abpeikar, S. (2022). Frontier-led swarming: Ro-

bust multi-robot coverage of unknown environments.

Swarm and Evolutionary Computation, 75:101171.

Wang, D., Wang, H., and Liu, L. (2016). Unknown

environment exploration of multi-robot system with

the fordpso. Swarm and Evolutionary Computation,

26:157–174.

Yamauchi, B. (1998). Frontier-based exploration using mul-

tiple robots. In Proceedings of the second interna-

tional conference on Autonomous agents, pages 47–

53.

Yanguas-Rojas, D. and Mojica-Nava, E. (2017). Explo-

ration with heterogeneous robots networks for search

and rescue. IFAC-PapersOnLine, 50(1):7935–7940.

Zhai, C. (2014). Sweep coverage of discrete time multi-

robot networks with general topologies. Kybernetika,

50(1):19–31.

Zhai, C. and Hong, Y. (2013). Decentralized sweep cover-

age algorithm for multi-agent systems with workload

uncertainties. Automatica, 49(7):2154–2159.

Zhang, C. and Noguchi, N. (2017). Development of a multi-

robot tractor system for agriculture ﬁeld work. Com-

puters and Electronics in Agriculture, 142:79–90.

Sweeping-Based Multi-Robot Exploration in an Unknown Environment Using Webots

255