Comparison Between Line-Followers and Free Movement Robots in Tasks Execution in A Simulated Environment

You might also like

You are on page 1of 6

2022 Latin American Robotics Symposium (LARS), 2022 Brazilian Symposium on Robotics (SBR), and 2022 Workshop on Robotics

in Education (WRE) | 978-1-6654-6280-8/22/$31.00 ©2022 IEEE | DOI: 10.1109/LARS/SBR/WRE56824.2022.9995776

Comparison between Line-Followers and Free Movement Robots in


Tasks Execution in a Simulated Environment
1st Igor Pardal Latini 2nd Wilson Eduardo Barioni
Universidade Tecnológica Federal do Paraná (UTFPR) Universidade Tecnológica Federal do Paraná (UTFPR)
Av. Sete de Setembro, 3165 Av. Sete de Setembro, 3165
Curitiba, Parana, Brazil Curitiba, Parana, Brazil

3rd Marco Teixeira 4th Flavio Neves-Jr


Universidade Tecnológica Federal do Paraná (UTFPR) Universidade Tecnológica Federal do Paraná (UTFPR)
Av. Sete de Setembro, 3165 Av. Sete de Setembro, 3165
Curitiba, Parana, Brazil Curitiba, Parana, Brazil

5th Lucia Valeria Ramos de Arruda


Universidade Tecnológica Federal do Paraná (UTFPR)
Av. Sete de Setembro, 3165
Curitiba, Parana, Brazil

lifting increases productivity and, in the long term, reduces


Abstract—In Industry 4.0, there is a constant pursuit for costs, with the so-called Industry 4.0 [2].
efficiency and productivity, and, searching to achieve this, the In this contexte, many factories use AGVs (Automated
usage of multiple robots to attend to several tasks simultaneously
is usual. Nowadays, there is plenty of robotic architectures, Guided Vehicles) not only for automatizing line production but
having their navigation system based on numerous different also for transportation of the by-products inside their facilities.
sensors. This work compares the line-follower robots, having a This automation also occurs in cargo stock, with warehouses
determined path, with the free-moving robots that navigate based using autonomous equipment to transport goods and carry out
on odometry, for load collecting and delivery tasks in specific other logistical processes like inventory management [3].
locations. For this, four simulated experiments are run in two
scenes, with two and four tasks, respectively. The comparison is Several issues must be addressed for the complete automa-
based on a difference in robot quantity, with more line-followers tion of entire distribution centers, including the number of
than free-moving ones, aiming to collect values such as idle time, robots to be used. Those centers cannot achieve the wished
task wait time, and the number of tasks done for each machine. efficiency and productivity by using only one robot. Using
The results show advantages in using less but more intelligent multiple autonomous machines makes it possible to attend to
robots with free movement. Such robots are not bound to one
unique task and can use their idle time to roam and complete numerous tasks simultaneously, accelerating the processes.
other processes. However, the task wait time is a concerning In real applications, several tasks are ready to be executed
issue, as too few robots for overcharged situations can increase at different times, which brings the question of how many
execution delays. robots a determined process needs. By assigning one robot
Index Terms—Robotics, simulation, line-follower robot, free to each task, we guarantee that the tasks run when they are
movement robot, fuzzy controller
ready, considering no planning failures and the time needed to
complete the tasks. With this approach, the downside is the idle
I. INTRODUCTION time of each machine since they are assigned to that specific
The crescent modernization of factories and the introduction task. Until the task is ready, that robot will stand idle, losing
of new logistic technologies induce a replacement of the hu- efficiency and therefore representing a cost to the company.
man labor, once indispensable for a company’s manufacturing In contrast, it is possible to use fewer robots attending to
processes, with robots [1]. Using machines to do the heavy multiple tasks. With no explicit path, the robot waits to receive
a call from the task and promptly goes to the location to
Universidade Tecnologica Federal do Parana (UTFPR), Av. Sete de Setem- execute it. In this way, fewer machines would be able to keep
bro, 3165, Curitiba, Parana, Brazil the same performance as the one task - one machine approach
Automation and Advanced Control System Laboratory (LASCA)
Graduate Program in Electrical and Computer Engineering (CPGEI) but with less idle time per equipment.
Wilson Eduardo Barioni eduardo.barioni@agvs.com.br For this, it is necessary to consider the navigation ap-
Igor Pardal Latini igorlatini@gmail.com proaches for load-carrying robots. Such approaches consider
robots with fixed paths, also called line-followers, and ones
978-1-6654-6280-8/22/$31.00 ©2022 IEEE with free movement. The first category could implement one of

145
Authorized licensed use limited to: BEIHANG UNIVERSITY. Downloaded on April 15,2023 at 01:37:11 UTC from IEEE Xplore. Restrictions apply.
many technologies. It can use an infrared ray sensor to follow a For the angular controller, the error is the difference between
line made with tape on the ground or use a magnetic sensor to the current angle of the robot and its desired angle facing the
follow a magnetic line [4]. The free movement robots are more target point. This can be achieved by Equation 3.
intelligent and use odometry from sensors like accelerometers,
gyroscopes, inertial measurement units, LIDAR, among others, θ = atan2(y2 − y1 , x2 − x1 ) (3)
to locate themselves [5]. Where θ is the angular error and atan2 is an arc-tangent
This work compares the usage of robots applying a line- function that takes the quadrants of (x, y) into consideration.
follower method to complete tasks with fewer robots utilizing (x2 , y2 ) is the target point, and (x1 , y1 ) the robot position.
a free movement strategy, regarding their idle times, the tasks’ An avoid obstacles Fuzzy Logic controller is designed and
wait times, and how many tasks each machine was able to switched from the PID when necessary. Fuzzy Logic allows
complete, using a simulated environment. representing uncertainty and vagueness. A fuzzy controller
The structure of this work is in such a way that Section uses linguistic variables and consists of three steps: fuzzify
II deals with the used methodology and concepts, while the inputs by associating them with the membership functions,
Section III contains aspects related to the development and apply the previously specified rules to calculate the outputs,
the simulation. In sequence, Section IV presents the results and de-fuzzify those outputs to get the final values [10].
obtained, concluding in Section V that also contains the future A TITO (Two Inputs and Two Outputs) fuzzy controller
work. is applied for obstacle avoidance control. The first input
II. M ETHODOLOGY is the distance to the closest obstacle, encompassing four
sets: “VeryClose”, “Close”, “Far” and “VeryFar”. The second
The comparison is carried out in a simulated environment
input is the angle to closest obstacle, encompassing five sets:
built into V-REP (Virtual Robot Experimentation Platform)
“VeryRight”, “Right”, “Center”, “Left” and “VeryLeft”. Figure
[6]. All codes are developed using Python language [7] and
1 shows the membership functions for both sets.
ROS (Robot Operating System) framework [8].
A. Criteria
The comparison between line-followers and free-movement
robots is based on their capabilities in completing tasks. Thus,
the criteria used for the comparison are idle time per robot,
maximum and average wait time per task, and the number of
tasks completed. The idle time is considered any time that
the robot is not doing a task nor in route to start one. The
wait time for the tasks is defined as the time passed between
that task being ready to be executed and one robot starting it. Fig. 1: Membership functions for distance and for angle to the
Other data, like traveled distance and route time for the free- closest obstacle (fuzzy controller inputs)
movement robots are also collected. These values, associated
with their respective experiments, can be seen in Section IV. For the outputs, the first one is the angular velocity, encom-
passing five sets: “VeryClock”, “Clock”, “Center”, “Anticlock”
B. Navigation and Control and “VeryAnticlock”. This output corresponds to the correc-
The free-movement robot utilizes two PID controllers [10] tion in the angular movement that should be applied so that the
to navigate around the scene. The first one controls linear robot can avoid the detected obstacle. Finally, the last output
speed and the other the angular speed. PID controllers are is the linear velocity, encompassing three sets: “VerySlow”,
mathematically given by “Slow” and “Normal”. The idea is, the closer the robot is to
Z t the obstacle, slower it will get, so it will have time to avoid
de(t) it. The membership functions of both outputs are shown in
u(t) = K p e(t) + Ki e(t)dt + Kd (1)
0 dt Figure 2.
Where u(t) is the control output, e(t) is the error between The rules generally make the robot motion slower the closest
the controlled variable and its set-point, K p is the proportional it is to the obstacle but also increase the angular velocity so
gain, Ki is the integral gain and Kd is the derivative gain. that the robot can easily do the deviant maneuver. Likewise, it
The linear speed error is calculated as the linear distance boosts the angular velocity if the obstacle is closer to the center
between the current robot position in the scene and the target of the robot from an angular standpoint. To avoid collision
point. The linear distance is the euclidean distance given by between two robots, the rule for a center obstacle is to “go
Equation 2. left”.
q III. D EVELOPMENT
d = (x2 − x1 )2 + (y2 − y1 )2 (2) A. Tasks
Where d is the linear distance, (x2 , y2 ) is the target point, The tasks comprehended collecting the load, going to
and (x1 , y1 ) is the robot’s current point. another location, and then delivering it, corresponding to a

146
Authorized licensed use limited to: BEIHANG UNIVERSITY. Downloaded on April 15,2023 at 01:37:11 UTC from IEEE Xplore. Restrictions apply.
Fig. 2: Membership functions for angular and linear velocity
Fig. 3: Robot used for line-follower and free-movement navi-
(fuzzy controller outputs)
gation approaches

standard transportation procedure. Each task has its defined


received in ROS. The maximum speed for each machine is
starting location, final location, and a fixed timer for when to
limited to 1m/s, keeping the two types of robots at the same
be ready to be executed.
levels.
The tasks are only representative since the loads’ physical
This study considers no-error odometry allowing the robots
characteristics and the load and unload difficulties are beyond
to know their location in the scene. In a real scenario, the
this study’s scope. For this reason, every task consists of
sensors would have inaccurate measurements, and the robot’s
these three steps: first, the robot waits a fixed amount of
positions could be unknown.
time to simulate collecting the load, then it goes to that
To effectively apply the values of linear and angular ve-
task’s previously determined destination, and finally remains
locities calculated by the PID or the Fuzzy controllers, it is
for another specified amount of time corresponding to the
essential to consider the robot’s kinematics model. The robot
load delivery operation. However, this definition has some
used here in has a differential system composed of two fixed
differences according to the type of robot that completes it.
wheels, in which the difference in speed between them causes
The line-followers robots always start in the fixed initial
the movement [13]. The angular and linear velocities to be
position of their assigned tasks. Since they have a line to
applied to the wheels are computed using Equations 4 and 5,
follow, their path is known from the beginning. With this,
.
every robot needs to return to its starting position after
delivering the load; thus, the tasks for these kinds of robots ωL vx
also consider this. For line-follower robots, the tasks have an vright = − + (4)
2r r
extra step of coming back to the initial position.
In contrast, the free movement robots do not have path ωL vx
vle f t =
+ (5)
constraints and are also not limited to any particular task. For 2r r
this reason, they do not need to return to their starting points. Where vright is the velocity applied to the right wheel and
However, they also have an extra step, but in the beginning, vle f t to the left wheel, ω is the angular velocity and vx the
since they have to reach the start point of the next task. linear velocity, both received through ROS topics. Finally, r
A scheduler must be created [11] to assign the tasks to free is the wheel radius and L the distance between the wheels.
movement robots. It checks the tasks, and, every time one
C. Timers
is ready, the scheduler puts it in a list and assigns it to the
idle robot closest to the starting point of such activity. This Using time measurements as criteria for comparison turned
assignment guarantees that only one robot performs a task at out to be problematic, especially considering the synchro-
a time, and if all robots are busy, the scheduler will not miss nism between the simulation and the Python scripts. Heavy
that call, keeping the activity in the list until one turns idle. processing and rendering can slow the machine down when
the simulation is running, making the time count wrong, as
B. Robot everything is slower than it should be. However, the scripts
The robot used in this work for line-follower and free- were using real-time from the ROS function, which made the
movement navigation approaches is the same. It is a simple loss of synchronism a real problem, with tasks being ready
one since it is an example. Any robot could be used if utilized while the robots were still on their way to completing them.
in conjunction with ROS. The robot can be seen in Figure 3. The use of simulation time instead avoids such a problem.
The robot consists of a cylindrical fairing, wheels with a The simulation time, as is shown in Figure 4, is an internal
0.1m radius and 0.25m distance between the wheels. It also variable in V-REP that is incremented by a value in each loop
has a LIDAR (Light Detection and Ranging) scanner on its top, of the simulation [14]. That means that, as the simulation
used for obstacle perception. This scanner is a Hokuyo URG slows down, this value will also slow its growth down, keeping
04LX UG01 [12], modified to use a 180 degrees arc range up. Therefore, while the values obtained as times do not
and increment of approximately 0.006 radians (0.34 degrees) correspond to seconds in any way, they can be used as a
with each beam, resulting in 512 measurements per message suitable means of keeping track of time since it is constant

147
Authorized licensed use limited to: BEIHANG UNIVERSITY. Downloaded on April 15,2023 at 01:37:11 UTC from IEEE Xplore. Restrictions apply.
between all parts, the simulation, and the scripts. For lack of a The second one, built in a 15m x 15m platform, represents a
better name, this unit of time will be referenced as “simulation more realistic situation, with more tasks and possible obstacles
time” or simply “st” in this work. in the path described by the cupboards. In this scene, there are
four tasks: Task 1 is the right one, Task 2 is the top one, Task
3 is on the left, and, finally, Task 4 is on the bottom side.
These nomenclatures will be used in Section III-E.
The paths, marked as black tape in both scenes, are only
used by the line-followers robots. The free movement ones
do not have any restrictions in their movements, but, in the
second scene, it is necessary to do some obstacle dodging.
Fig. 4: Simulation time loop [14]
E. Experiments
In total, four experiments were run, one with the first scene
D. Scenes and three with the second scene. As seen in Section III-C,
The simulated environment has two scenes created using V- the values presented are in “st”, and do not correspond to real
REP (Virtual Robot Experimentation Platform). The first one seconds.
is a simple scene to serve more as a “proof of concept”, while For the first experiment, the values used by the tasks to
the second scene is more complex and offers more challenges be ready are presented in Table I. The “Initial time” can be
for the robots, being a light simulation of a possible space interpreted as if the time count already starts at this value,
in a warehouse. The first scene is shown in Figure 5, while and “Time to be ready” is the target count in which the task
the second one is present in Figure 6. Although both scenes will be ready to be completed. As soon as the task is ready,
illustrate the line-following robots, the scenes with the free the timer goes to zero and another counting is initiated. This
movement ones are the same, with the robot starting in the means that the initial timer, as the name says, only counts the
center. first call of the task. This experimentran for 830st.

TABLE I: Task times for the first experiment

Task Initial time (st) Time to be ready (st)


Task 1 0 100
Task 2 50 100

Table II presents the values for the second experiment using


the second scene. In this experiment, there are four tasks, two
more than in the previous analysis. Two of those tasks are also
longer, taking more time to complete. This experiment ran for
Fig. 5: First scene 700st.
TABLE II: Task times for the second experiment

Task Initial time (st) Time to be ready (st)


Task 1 320 340
Task 2 0 100
Task 3 0 240
Task 4 0 150

The third experiment, which also uses the second scene,


was made to stress the free movement robot, reducing the time
necessary for the tasks to be ready. The values are present in
Fig. 6: Second scene Table III. This experiment also ran for 700st.

TABLE III: Task times for the third experiment


The first scene is a 10m x 10m platform with no obstacles.
The two tasks correspond to the two lines, as seen in the figure. Task Initial time (st) Time to be ready (st)
The one on the right is Task 1, while the one on the left is Task 1 165 175
Task 2. The robots are positioned in the initial location, ready Task 2 0 60
Task 3 0 130
to start collecting the load when their respective tasks turn Task 4 30 60
ready.

148
Authorized licensed use limited to: BEIHANG UNIVERSITY. Downloaded on April 15,2023 at 01:37:11 UTC from IEEE Xplore. Restrictions apply.
In the fourth and last experiment, another free movement next ones, maintain the same format as the previous ones but
robot was added, seeking to decrease the wait time for the are in a more concise format, using abbreviations to facilitate
tasks. The initial location of both is right at the center, with the view.
robot one dislocated 3 meters up and robot two 3 meters down. Here, we can see the same pattern of the first experiment.
This experiment used the same time values as experiment Even after adding 275.75st equivalent to the route time (re-
three, seen in Table III, and was also run for 700st. sulting in 365.7st), it represents only 22.9% of the idle time
of the line-followers, indicating that there is too much free
IV. R ESULTS
time for them. Regarding the task times, the relatively high
Table IV contains the results for experiment 1, having the difference between the maximum wait time and the average
idle time, the number of tasks completed and the traveled one can be explained by the many different paths the free
distance for each robot, as well as a sum for easy comparison movement robot can take to each task, which sometimes can
between all line-followers and the one free movement. Table be closer than others. Also, the situation of a task being ready
V contains the maximum and average wait time for each while the robot is still finishing another task is much more
task and the average time to complete the task. To make the present here. However, no process is unattended; hence the
tables easier to read, the line-follower robots are abbreviated number of completed tasks is the same.
to “Line”, while the free movement ones are abbreviated to
“Free”. TABLE VI: Results for the second experiment
TABLE IV: Results for the first experiment Line 1 Line 2 Line 3 Line 4 Total Free
Idle(st) 368.15 405.9 366.45 453.25 1593.75 86.95
Line 1 Line 2 Total Line Free Comp. 2 6 2 4 14 14
Idle time (st) 458.65 409.35 868 97.85 Dist.(m) 81.9 69.87 46.28 46.67 243.72 215.62
Tasks completed 7 8 15 15
Travelled distance (m) 88.08 93.13 181.21 229.53

TABLE VII: Task related results for the second experiment


The route time corresponds to the time the free robot
spent going to the initial location of each task. This time is Task 1 Task 2 Task 3 Task 4
independently counted, and its value is equal to 314.4st. The Max. wait (st) 53.7 56.8 69.54 52.35
Avg. wait (st) 35.85 23.92 43.64 34.16
addition of the route and the idle times resulted in 412.25st, Avg. comp. Line (st) 165.92 49.02 113.75 48.95
which is still only 47.5% of the total time for the line- Avg. comp. Free (st) 33.05 24.87 24.85 24.87
followers. This means that one free movement robot could, in
this situation, quickly do the job of two line-followers robots
For experiment three, the free movement robots are under
and would be idle nearly half the time. The total number of
stress by decreasing the time necessary for each task to be
completed tasks is the same, indicating that no process was
ready, pushing it almost to the limit. The results can be seen
left unattended or not terminated.
in Table VIII, with the task related result in Table IX. Since
TABLE V: Task related results for the first experiment this is the same scene as experiment two, the values for average
completion time are similar, thus they are removed from the
Task 1 Task 2 table.
Max. wait time (st) 20.25 20.25 The route time in this experiment is 293.25st, which, added
Avg. wait time (st) 20.19 19.15
Avg. completion time Line (st) 48.76 48.7 to the idle time for the free movement robot, results in 303.85.
Avg. completion time Free (st) 27.19 27.21 Here, the free movement equipment is clearly insufficient to
attend to all tasks, having done only 15 compared to the 28
done with the line-followers. The idle time presented, 10.6st,
Concerning the tasks, it is possible to see that the maximum
is for the first task to be up. After that, there was no more idle
wait time and the average one are very close, indicating that
time for this robot. This shows the importance of looking at
this value comes from the robot going to the initial position
both the idle time and the number of tasks completed.
of the tasks. The difference between the completion times is
Comparing Table IX with Table VII, the high wait values
explained in Section III-A, but is related to the time to go back
show that the single robot is unable to attend to the four tasks
to the initial position on the line-followers. It is important to
using these new, lower refresh times.
note that the completion time for the line-follower is relative
to each robot, as they can only complete one task. TABLE VIII: Results for the third experiment
It is important to note that, since the line-follower robots
are limited to each task, they have no wait time, and this value Line 1 Line 2 Line 3 Line 4 Total Free
only applies to the free movement equipment. Idle(st) 37.5 169.6 192.15 150.9 550.15 10.6
Comp. 3 10 4 11 28 15
The results for the second experiment are shown in Table Dist.(m) 163.61 127.05 103.79 130.48 524.93 241.17
VI, reflecting the increase in complexity. Table VII presents
the results from a task standpoint. Those tables, and also the

149
Authorized licensed use limited to: BEIHANG UNIVERSITY. Downloaded on April 15,2023 at 01:37:11 UTC from IEEE Xplore. Restrictions apply.
TABLE IX: Task related results for the third experiment and other approximations. In a real-world setting, the behavior
observed could differ from the one presented here since several
Task 1 Task 2 Task 3 Task 4
Max. wait (st) 229.9 303.75 291.2 354.05
variables are not considered in a simulation.
Avg. wait (st) 124.1 166.46 218.63 202.64
R EFERENCES
[1] A. Berg, E. F. Buffie and L. Zanna, “Robots, growth, and inequality”.
Finance & Development, vol. 53, n. 3, 2016, pp. 10-13.
In the last experiment, we add a second free movement [2] S. Vaidya, P. Ambad and S. Bhosle, “Industry 4.0–a glimpse”. Procedia
robot. The results are shown in Table X, while Table XI manufacturing, vol. 20, 2018, pp. 233-238.
[3] P. Baker, Z. Halim, “An exploration of warehouse automation implemen-
contains task related results. The tables show that it is possible tations: cost, service and flexibility issues”. Supply Chain Management:
to fix the situation presented in experiment three using two An International Journal, 2007.
free-movement robots. However, the idle times indicate that [4] K. M. Hasan, Abdullah-Al-Nahid and A. Al Mamun, “Implementation
of autonomous line follower robot”, 2012 International Conference on
there is not much room left. The total number of tasks suggests Informatics, Electronics & Vision (ICIEV), 2012, pp. 865-869, doi:
that the free movement robots are keeping up with the line- 10.1109/ICIEV.2012.6317486.
followers, but the maximum wait time presented at Task 4, [5] S. A. S. Mohamed, M. -H. Haghbayan, T. Westerlund, J. Heikkonen,
H. Tenhunen and J. Plosila, “A Survey on Odometry for Autonomous
112.85st, shows that there are some times that the tasks Navigation Systems”, in IEEE Access, vol. 7, pp. 97466-97486, 2019,
accumulate. Even so, this experiment shows that the free doi: 10.1109/ACCESS.2019.2929133.
movement robots can mostly handle the situation at almost [6] COPPELIA ROBOTICS, “Coppelia Sim”. Available at:
https://www.coppeliarobotics.com/coppeliaSim. Acessed in: May
full capacity. 18, 2022.
[7] PYTHON SOFTWARE FOUNDATION, “Python Language
TABLE X: Results for the fourth experiment Site: Documentation, 2020”. Documentation page. Available at:
https://www.python.org/doc/. Accessed in: May 18, 2022.
Line 1 Line 2 Line 3 Line 4 T. Line Free 1 Free 2 T.Free [8] OPEN ROBOTICS, “ROS - Robot Operating System”. Available at:
Idle(st) 37.5 169.6 192.15 150.9 550.15 75.95 67.8 143.75 https://www.ros.org/. Accessed in: May 18, 2022.
Comp. 3 10 4 11 28 14 14 28 [9] M. Quigley, et al., “ROS: an open-source Robot Operating System.”
Dist.(m) 163.61 127.05 103.79 130.48 524.93 217.23 251.8 469.03
ICRA workshop on open source software, vol. 3. No. 3.2. 2009.
[10] K.M. Passino, “Fuzzy control”, Addison-Wesley, 1998.
[11] M. Song, T. -J. Tarn and N. Xi, “Integration of task scheduling, action
planning, and control in robotic manufacturing systems”, in Proceedings
TABLE XI: Task related results for the fourth experiment of the IEEE, vol. 88, no. 7, 2000, pp. 1097-1107, doi: 10.1109/5.871311
[12] HOKUYO, “Scanning Laser Range Finder URG-04LX-UG01”. Avail-
Task 1 Task 2 Task 3 Task 4 able at: https://hokuyo-usa.com/application/files/5115/8947/8197/URG-
Max. wait (st) 65.1 82.7 60.3 112.85 04LX-UG01 Specifications Catalog.pdf. Accessed in: May 18, 2022.
Avg. wait (st) 43.44 39.47 47.35 48.95 [13] S. K. Malu and J. Majumdar. “Kinematics, Localization and Control
of Differential Drive Mobile Robot”, Global Journal of Research In
Engineering, [S.l.], 2014. ISSN 2249-4596.
[14] COPPELIA ROBOTICS, “Simulation”. Available at:
V. CONCLUSIONS https://www.coppeliarobotics.com/helpFiles/en/simulation.htm.
Accessed in: May 18, 2022.
The result analysis points to the advantage of using fewer,
more intelligent robots to do tasks instead of just assigning a
line-follower robot to each process. By having free-movement
robots, it is possible to take advantage of their idle time to
move them to other places and keep them busy, maximizing
the investment to acquire it and do the necessary infrastructure
changes.
However, it is important to keep the wait time of each task
in mind. An inherent delay is always associated with the free
movement robot since it must drive through the initial place
of each process. Also, the time each task takes to become
available is critical, as it can overload the free movement
robots, as seen in experiment three.
For future works, other types of navigation systems will
be compared, such as robots running with only a Fuzzy
controller, without PID. Also, the complexity of the scenes
can be increased, adding more robots, more tasks, dynamic
obstacles, etc. Moreover, the scheduler can consider aspects
other than only distance. Finally, the power supply can be also
analyzed, simulating the consumption and creating a region
and a wait time for recharging.
As a final consideration, this work has used a simulated
environment, including measurements from free-error sensors

150
Authorized licensed use limited to: BEIHANG UNIVERSITY. Downloaded on April 15,2023 at 01:37:11 UTC from IEEE Xplore. Restrictions apply.

You might also like