Research Article
Open Access
Multiple Mobile Robots Navigation and
Obstacle Avoidance Using Minimum Rule
Based ANFIS Network Controller in the
Cluttered Environment
Anish Pandey* and Dayal R. Parhi
Mechanical Engineering Department, N.I.T, Rourkela, Orissa, India
*Corresponding author: Anish Pandey, Mechanical Engineering Department, N.I.T, Rourkela, Orissa, India, E-mail:
@
Received: January 18, 2015; Accepted: January 27, 2016; Published: February 10, 2016
Citation: Pandey A, Parhi DR (2016) Multiple Mobile Robots Navigation and Obstacle Avoidance Using Minimum Rule Based ANFIS
Network Controller in the Cluttered Environment. Int J Adv Robot Automn 1(1): 1-11. DOI:
http://dx.doi.org/10.15226/2473-3032/1/1/00102
Abstract
The motion control problem of an autonomous wheeled mobile
robot has been widely investigated in the past two decades. In this
article, the minimum rule based Adaptive Neuro-Fuzzy Inference
System (ANFIS) controller has been presented for the safe navigation
of single and multiple mobile robots in the cluttered environment
by using the sensor-based steering angle control technique. The
ultrasonic range finder sensor and sharp infrared range sensor
have been used to read the front, right and left obstacle distance.
This obstacle distance information is fed to the input of the ANFIS
controller for selecting the suitable steering angle to achieve the
collision-free path while mobile robots are moving to reach the
goals in the cluttered environment. The advantages of this ANFIS
network controller have a rapid response and can make the robots
react efficiently in the real environment. The simulation result
shows that the proposed ANFIS controller gives the safe, optimal and
smooth travelling path in the cluttered environments. Moreover, the
simulation results are compared with experimental results in the real
environment to prove the authenticity the controller.
Keywords: Navigation; Sensors; Cluttered environment;
Obstacle; Steering angle
Introduction
The present article explores the application of ANFIS network
controller to solve the path planning problem of the multiple
mobile robots to move from the current position to the goal
position in the cluttered environment. The simulation and the
experimental result show that the proposed ANFIS controller
improves navigation performance in the unknown cluttered
environments. ANFIS is a major topic in the field of neural and
evolutionary soft computing. This algorithm plays a significant
role in the designing of an intelligent autonomous control system.
ANFIS makes the mapping between the input and output using
the neural network and fuzzy logic for the better performance.
This ANFIS architecture has been adapted from Matlab software
package. ANFIS is a combination of fuzzy logic and Artificial Neural
Networks (ANN), where the fuzzy logic is capable of handling
the linear and non-linear uncertainty, or both. The ANN is used
to tune the parameters of the input and output membership
functions [1]. ANFIS may be used as linear or non-linear, or both
type dynamic systems, which solves the real system problem by
using empirical dataset (experimental or predicted).
The ANFIS is the product of two methods, neural networks,
and fuzzy systems. If both these intelligent methods are combined,
better reasoning will be obtained in term of quality and quantity.
In other words, both fuzzy reasoning and neural network
calculation will be available simultaneously [7]. This ANFIS
technique has been successfully applied by many researchers for
sensor-based autonomous control mobile robots in the different
environment. To address the best and optimal path, the robots
required to move with suitable path planning algorithm, which
calculates the minimum path length between any two points
i.e. robot to obstacles or robot to the goal [3]. Pothal and Parhi
[8] have proposed the sensor based Adaptive Neuro Fuzzy
Inference System (ANFIS) controller for the navigation of single
and multiple mobile robots in the highly cluttered environment.
The authors have tried to design a control system architecture,
which avoids the obstacle autonomously and reaches the target
efficiently in all types of environments.
Hu and Brady [9] have described the design and
implementation of parallel processing control architecture for
sensor-based real-time local navigation of the mobile robots. The
authors have done many experiments to prove the authenticity
of the system. Contreras-Cruz, et al. [10] have proposed the
evolutionary approach and artificial bee colony algorithm to solve
the mobile robot path planning problem. Parhi and Mohanta [11]
have developed the path planning of the multiple mobile robots
in an unknown cluttered environment using a Petri-potential
fuzzy hybrid controller with different membership functions.
Furthermore, the set of sensors has attached to the robots to
detect the nearby obstacles of the robot, to protect the robot from
inter-collision.
Mohanta, et al. [12] have integrated the Genetic Algorithm
(GA) and Petri-Net technique for safe and collision-free path
planning of the multiple robots with multiple targets in the
cluttered environment. The Genetic Algorithm (GA) is used for
finding an optimal or near optimal best paths in the cluttered and
complex environments. Pradhan, et al. [13] have designed the
navigation techniques for one thousand mobile robots using a
fuzzy logic controller with different membership functions in an
unknown environment.
Hu, et al. [14] have designed the real-time path planning
of a mobile robot by using the sensor-based neural network
technique. The different sensors such as sonar sensors and
infrared range finder sensors are used to control the steer of
the robot to protect against unexpected hurdles in the uncertain
terrain. Boubertakh, et al. [15] have prepared a new simple fuzzy
logic eight rule-based controllers for mobile robot navigation,
inspired by the human knowledge and used for obstacle avoidance
and goal seeking behaviour. Kubota, et al. [16] have made the
behaviour-based fuzzy controller for mobile robot navigation
and collision avoidance in the non-stationary environment, and
the conventional genetic algorithms (GAs) are integrated with for
path optimization.
Hoy, et al. [17] has studied the navigation problem of
multiple vehicles through an unknown static environment with
limited sensing and communication. Begum, et al. [18] has
done the hybridization of fuzzy logic and genetic algorithm for
solving the localization and mapping problem of several mobile
robots. Korayem, et al. [18] have investigated the dynamic
modelling and optimal point-to-point motion planning of a nonholonomic
mobile robot in cluttered environments. Ahmadzadeh
and Ghanavati [20] have proposed the PSO algorithm based
navigation method for multiple mobile robots.
In this work, we propose the path planning and obstacle
avoidance for multiple mobile robots in the cluttered environment
using ANFIS network controller. Motivated by the aforementioned
researchers, the primary objective of this article is to show how
to guide the multiple autonomous mobile robots in an unknown
cluttered environment using the minimum eight-rule based
ANFIS controller. The proposed ANFIS network system is used
to adjust the motions, directions and movements of the mobile
robot to reach the goals with obstacle avoidance strategy. When
the robots near to the obstacle, the obstacle avoidance behaviour
a reactive, otherwise the goal finding behaviour will continue.
The developed controller receives the input data Front Obstacle
Distance (FOD), Right Obstacle Distance (ROD), and Left Obstacle
Distance (LOD) from the various sensors. These sensor signals
are fed to the input to the ANFIS network to provide the output
Steering Angle (SA) control commands for the mobile robots. The
proposed ANFIS controller trains the mobile robot system online
dynamically. This ANFIS controller has three inputs and a single
output. The input has two generalized bell-shaped (gbellmf)
membership function, and output has Sugeno-type constant
membership function. This paper is divided as follows. Section
1 presents the introduction. The dynamic modelling of the
mobile robot is given in Section 2. ANFIS network architecture for multiple mobile robots navigation is proposed in Section 3.
Section 4 shows the simulation result and discussion and also
a comparison with previous works to the proposed method.
Section 5 represents the experimental setup and its results and
discussion for validating the proposed ANFIS controller. Finally,
in Section 6, conclusions are presented.
Dynamic Modeling of the Mobile Robot
This work has been developed for wheeled type mobile robot,
the front wheels of the robot are attached to the separate motors
responsible for direction change (steer control), and the rear one
is a caster wheel for balance. The movement of the mobile robot
has been controlled by the steering (motor speed) of the front
wheels. The dynamic modelling of the robot in the environment
is shown in Figure 1. In Figure 1, L, VR and VL denotes the track
width, right wheel velocity, and left wheel velocity of the mobile
robot, respectively. The point P is located between the centres
of the driving wheels axis. The landmark (O, X, Y) is related to
the field navigation environment. The central point (Xc, Yc) is
the middle of the two front steering wheels, and θ is the angle
of steering about the axis (O, X).The following dynamic equation
describes the motion and orientation of the robot in the field
navigation environment: -
Where j = 0, 1, 2,…, n.
Where X [j] is the current position of the mobile robot in the
X-axis, and X [ j +1 ] is the updated position of the robot when it
moves forward by using Equation 1. Y [j] is the current position
Figure 1: The dynamic modeling of a mobile robot.
of the mobile robot in the Y-axis, and Y [ j +1 ] is the updated
position of the robot when it moves forward by using Equation
2. The angle θ is the steering angle of the mobile robot, and this
steering angle varies (or updated) according to the position of
a goal in the environment. The Gx and Gy are the goal (or target)
point in the X and Y-axis, respectively. The V is the centre velocity
in meter per second (m/sec), which depends upon the Diameter
(D) of the wheel in meter, and the Angular Velocity (N) of the
motor in RPM (Rounds Per Minute). The maximum and minimum
velocity of the mobile robot used for navigation is 16.7cm/sec,
and 6.7 cm/sec respectively.
The following conditions are used to control the motion and
orientation (steering angle) of the mobile robot: -
If < VL VR , then the robot turns left side Figure 4
If > VL VR , then the robot turns right side Figure 5
If = VL VR , then the robot moves straight Figure. 6
ANFIS Network Architecture for Multiple Mobile
Robots Navigation
Controller Architecture
The ANFIS controller for the multiple mobile robots navigation
can be easily implemented for real-time applications because
this controller handles unknown and uncertain situations with
the simple computations. This navigation controller is working
based on the local information collect from the different sensor.
Figure 2: Matlab based algorithm to tune the parameters.
Figure 3: The architecture of mobile robots navigation based on ANFIS
control system.
Figure 4: Positive steering angle.
Figure 5: Negative steering angle.
Figure 6: Zero steering angle.
For the safe navigation without any human intervention, the
mobile robot should receive information about its surroundings
environment by using the array of sensors. The proposed ANFIS
controller has three inputs and a single output, the controller
receives input (obstacle distance) from the ultrasonic range
finder sensors and sharp infrared range sensor and provides the
correct steering angle as an output. The training of this ANFIS
model has been done by using fifteen data set given in Table 1.
The network is trained by fifteen specially designed sample data
to deal the different behaviours. The ANFIS controller uses the set
of sensors to detect the obstacle distance (front, right, and left)
information as inputs and the steering angle as an output varies
according to the obstacle distance data. Table 1 is described here,
if the robots are finding obstacles at a distance 20cm to the front,
20cm to the right, and 100cm to the left, then the mobile robot
turns 74.3 degrees (left side or anticlockwise direction) means
positive steering angle to reach the goals with obstacle avoidance
Figure 4. If the robots are finding obstacles at a distance 20cm
to the front, 100cm to the right, and 20cm to the left, then
the mobile robot turns-65.9 degrees (right side or clockwise
direction) means negative steering angle to reach the goals in an
environment with obstacle avoidance Figure 5. If the robot finds
an obstacle at a distance 50cm to the front, 15 cm to the right, and 15 cm to the left, then the mobile robot moves straight means
zero degree steering angle to reach the goals in an environment
with obstacle avoidance Figure 6. If there are no obstacles near
to the robot, goal finding behaviour becomes active. The negative
steering angle means the right wheel velocity will be low, and the
left wheel velocity will be high respectively, for ANFIS sensorbased
control of multiple mobile robots navigations in the
cluttered environment. The positive steering angle means the
right wheel velocity will be increased much more than the velocity
of the left wheel, for ANFIS sensor-based control of multiple
mobile robot navigations in the cluttered environment. The
zero steering angles means the both wheel velocities will move
at the same speed, for ANFIS sensor-based control of multiple
mobile robot navigations in the cluttered environment. The
navigation algorithm plays a sensory act control cycle where the
mobile robot senses the nearby obstacle to make the appropriate
decisions in the every iteration. This ANFIS network controller
collects the data from the array of sensors to control the speed of
the robot according to the obstacle detection. Thus, the controller provides the directional control of mobile robots, planning of
its path to the reach goals, tracking a collision-free optimum
travelling path in the environments. The controller has been
designed by the Matlab ANFIS toolbox [4], and its programming
is also done by the author. Their descriptions are shown in Figure
2 (Matlab based algorithm to tune the parameters). Figure 3
shows the architecture of our incorporated ANFIS control system
to achieve the movement and direction (steering angle) control
for the mobile robots.
Table 1: Predicted training data for ANFIS network based control for multiple mobile robots navigation.
S. No. |
F.O.D.
'cm' |
R.O.D.
'cm' |
L.O.D.
'cm' |
S.A.
'degree' |
TURNING DIRECTION |
1. |
20 |
20 |
100 |
74.3 |
LEFT |
2. |
20 |
100 |
20 |
-65.9 |
RIGHT |
3. |
100 |
100 |
10 |
-70.4 |
RIGHT |
4. |
25 |
50 |
75 |
55 |
LEFT |
5. |
40 |
60 |
100 |
59.4 |
LEFT |
6. |
15 |
100 |
100 |
72.8 |
LEFT |
7. |
25 |
100 |
50 |
-22.9 |
RIGHT |
8. |
15 |
15 |
15 |
73.4 |
LEFT |
9. |
50 |
15 |
15 |
0 |
STRAIGHT |
10. |
20 |
10 |
10 |
77 |
LEFT |
11. |
100 |
15 |
15 |
0 |
STRAIGHT |
12. |
25 |
10 |
10 |
77.2 |
LEFT |
13. |
100 |
100 |
15 |
-70.5 |
RIGHT |
14. |
100 |
25 |
20 |
0 |
STRAIGHT |
15. |
100 |
100 |
100 |
-70.4 |
RIGHT |
The Training Methodology
ANFIS is a method of formulating the relationship between the
input and the output using the fuzzy logic mathematical concept.
There are the two types of Fuzzy Inference Systems (FIS) that
can be classified as Mamdani-type and Sugeno-type. The ANFIS
does work based on the Sugeno-type neuro-adaptive learning
technique. This technique has more compact and computationally
efficient, and its capability to customize the membership
functions compare to the Mamdani-type fuzzy system. The
adaptive techniques can be used to generate the membership
function and rules automatically so that the fuzzy system delivers
the best result from dataset [2]. In this paper, Adaptive Neuro-
Fuzzy Inference System (ANFIS) is used, whose output is defined
as the constant type. The final output of the ANFIS controller
is calculated by the weighted average method. This proposed
ANFIS controller has three input variables x (F.O.D.), y (R.O.D.),
and z (L.O.D.), and a single output variable f (S.A.), which is
connected to each other by the fuzzy rules. The architecture of
a typical ANFIS has three inputs, eight rules and a single output.
The output is defined by the first-order function. The each input
has two generalized bell-shaped (gbellmf) membership functions
(MFs). Generally, the ANFIS controller works based on the firstorder
Sugeno fuzzy model in the following form: -
Where i =1, 2,…,8, and the symbols pi , qi , ri
and si are the
coefficients of output membership functions fi with the eight
rules. Mi , Ni and Oi are the membership functions (MFs) of
inputs x, y and z respectively of the fuzzy sets.
The Matlab software package is used to design the ANFIS
architecture. The input layer addresses the Front Obstacle
Distance (FOD), Right Obstacle Distance (ROD), and Left Obstacle
Distance (LOD). The output layer addresses the Steering Angle
(SA). In this architecture, the inputs are directly connected to the
input membership functions (input mf); the input membership
functions are connected to the rules; rules are connected to the
output membership functions (output mf), and finally this output
membership function is connected to the output. The input
and output have three neurons and single neuron respectively.
For training the ANFIS network, the TRAIN function of the
Matlab software package is used. This function works on ANFIS
hybrid learning algorithm, which is the combination of forward
and backward neural network method. The proposed ANFIS
architecture has five layers Figure 7, which perform the different
actions for the controller. Their details are given below: -
Layer 1: The first layer is an adaptive node. It generates
membership grade for the inputs. The outputs of this layer are
defined as: -
Figure 7: The general architecture of the ANFIS network.
Where n =1, 2,…,8, and μMn , μNn and μOn are the input
membership functions that can be triangular, trapezoidal,
gaussian, generalized bell-shaped and other shape functions. In
this paper, the following generalized bell-shaped membership
function is used: -
The similar generalized bell-shaped membership equation
will be written for μNn and μOn Where a, b and c are the premise
parameters of the generalized membership function; called as
the half width, slope control, and centre respectively.
Layer 2: The nodes in this layer are also called as the rule
layer, indicating that they perform a simple multiplier. The
outputs of this layer are represented as: -
The Wn represents a firing strength or the truth value, of the
nth rule and n =1, 2, 3…8 is the number of Sugeno-type fuzzy
rules.
Layer 3: It is also called as the normalization layer, which
plays a normalization role in the ANFISnetwork. The outputs of
this layer can be represented as: -
Layer 4: Every node in this layer is called adaptive node,
whose output is simply the product of the normalized firing
strength. A defuzzification node determines the weighted
consequent value of a given rule. The connection between the
inputs and output of this layer can be expressed as follows: -
Layer 5: It is represented by a single summation of all incoming
nodes. This single node is a fixed node, which determines the sum of the outputs of all defuzzification nodes and gives the overall
system output that is steering angle as given below: -
The structure of ANFIS controller has two membership
functions for all inputs, i.e. front, right and left obstacles. These
inputs are specified in the form of generalized bell-shaped
membership function. Because this generalized bell-shaped
membership function covers a wide area, and, therefore,
it delivers better results compared to other functions. The
proposed ANFIS controller generates the rules and tunes the
correct membership function grade from the dataset. During the
learning process for the safe navigation of multiple mobile robots
in the cluttered environment, the ANFIS modifies the inputs
and output membership function parameters with the objective
of minimizing the sum of square error of the output (Steering
Angle). The training dataset is used to train the ANFIS network
controller while the testing dataset is used to verify the accuracy
and effectiveness of the trained ANFIS network controller for the
computation of the data quality evaluation. The specifications of
the proposed ANFIS network controller have been given in Table
2. The proposed ANFIS controller architecture is verified through
the mean squared error (M.S.E.) method: -
Where SAa is the actual steering angle, SAp is the predicted
steering angle value, and k is the number of observations.
Simulation Results and Discussion
This section describes the successful simulation results using
ANFIS controller in the various environments. Simulation result
shows the method can be used for wheeled mobile robots moving
Table 2: The specification of the proposed ANFIS controller.
Name |
Specification |
Generate FIS Type |
Sugeno |
The Initial FIS Model |
Grid Partition |
Decision Method for Fuzzy Logic Operation AND (Minimum) |
Product (Prod) |
Decision Method for Fuzzy Logic Operation OR (Maximum) |
Probabilistic (Probor) |
Output Defuzzification Method |
Weighted Average (wtaver) |
Number of Membership Functions (nummfs) for F.O.D. |
2 |
Number of Membership Functions (nummfs) for R.O.D. |
2 |
Number of Membership Functions (nummfs) for L.O.D. |
2 |
Membership Function Type (mftype) |
Generalized Bell-Shaped |
Number of Rules |
8 |
Output Membership Function |
Linear |
Train FIS Optimization (optim) Method (Learning Algorithm) |
Hybrid |
Number of Epochs (numepochs) |
500 |
in the cluttered environment. The sensor data are used to control
the mobile robots during simulation for the obstacle avoidance.
Furthermore, the application of this ANFIS controller for mobile
robots navigation has been also discussed. The simulation results
in Figures 8-11 exhibit that the mobile robots start moving from
start position avoid the many obstacles in the environment to
reach goal positions with the possibility of minimum travelling
path. The proposed navigation strategy is tested by simulating the
three navigation actions such as obstacle avoidance, robots intercollision
avoidance, and goal-seeking in the environment. The
path traced by the single and multiple robots in the environment
among the various obstacles using minimum rule based ANFIS
network controller has been presented in Figures 8-12. The
Figure 8 (i) and (ii) shows the navigation of single mobile robot in
the different environments; Figure 9showsthe navigation of two
mobile robots, Figure 10 shows the navigation of three mobile
robots and Figure 11 shows the navigation of four mobile robots
respectively. Total travelling path lengths and time taken to
reach the goals in the both simulation and experiment modes are
measured in centimetre by using statistical method for one, and
two robot(s) respectively, and it is given in Table 3. Minimise the
travelling path length and search time is the primary objectives
for the mobile robots navigation in different environments.
From the simulation results, it is clear that, the developed ANFIS
controller can drive the robots smoothly without collision in the
cluttered environment. Moreover, the experimental results of
these simulations have been shown in Figures 16-17
Comparison with Previous Works
This section describes the simulation result comparisons
between the Zhang, et al. [5], Montaner, et al. [6] models and the
proposed minimum rule based ANFIS network controller. The
performance of this ANFIS controller is evaluated on the basis of
the following criteria:-
(a) Travelling path length
(b) Travelling path smoothness
Figure 8: Obstacle avoidance in the different cluttered environment by
the single mobile robot using minimum rule.
Figure 9: Obstacle avoidance in theenvironment by two mobile robotswith
two goals using the minimum rule based ANFIS network controller.
Figure 10: Obstacle avoidance in the environment by three mobile
robots with three goals using the minimum rule based ANFIS network
controller.
Figure 11: Obstacle avoidance in the environment by four mobile robots
with four goals using the minimum rule based ANFIS network controller.
Figure 12: The simulation results comparison between the (i) Zhang
et al. [5] model and the (ii) ANFIS controller in the same environment.
Figure 13: TThe simulation results comparison between the (i) Montaner
et al. [6] model and the (ii) ANFIS controller in the same environment.
Figure 14: The simulation results comparison between the (i) Montaner
et al. [6] model and the (ii) ANFIS controller in the same environment.
Figure 15: Experimental mobile robot.
Zhang, et al. [5] have developed the reactive fuzzy logic based
control strategy for mobile robot navigation. The authors have
integrated this fuzzy logic controller with the RAM based neural
network to decrease the number of fuzzy rules. Where the fuzzy
rule-based controller are used to interpret sensory information
and the neural network provides the appropriate actions to
avoid obstacles and does control the heading angle of the robot.
The four sharp infrared (IR) range sensors are mounted on the
left, right, left-front and right-front side of the mobile robot to
detect the obstacles. The two stepper motors are used to drive
the each rear wheel to facilitate turns or forward and backward
movement.
Montaner, et al. [6] have designed the fuzzy logic controller for autonomous mobile robot path planning, which reduces
the navigation time from a start position to the end position
in an environment. They developed a fuzzy controller based
experimental mobile robot that establish a map between input
space (information receiving from ultrasonic sensors) to
control the direction and velocity (output space) in real time
autonomously.
The performance of this proposed ANFIS controller is mainly
evaluated by navigation path length and path smoothness.
Figures 12-14 show the simulation comparisons between the
ANFIS controller and the previous models [5-6]. Table 4 and 5
show the path traced by the robot using the ANFIS controller and
the previous models [5-6] in the various environments. In Table 4, the path traced by the robot has been found 91.3cm and 78.9cm
by using Zhang, et al. [5] model and the current ANFIS controller,
respectively in the same environment. Similarly, in Table 5 ,the
path covered by the robot has been found 100.7cm and 95.2cm
and so on, by applying Montaner, et al. [6] model and current
ANFIS controller, respectively in the same environment. From
the simulation comparisons, it is clearly seen that, the developed
ANFIS controller efficiently drives the robots in the same path
planning problem.
Table 3: Navigation path lengths by the robots during simulation and experiment.
Figures
description |
Navigation path length in 'cm' |
Time taken to reach the goals in 'second' |
Simulation result |
Experimental result |
Simulation result |
Experimental result |
Fig.8 (i) and Fig.16 Single mobile robot |
107.2 |
115.3 |
9.4 |
10.2 |
Fig.9 and Fig.17 Two mobile robots |
103.5 and 57.1 res. |
111.7 and 61.8 res. |
9.2 and 5.1res. |
9.9 and 5.5res. |
Table 4: The Results of path traced by the robot using the ANFIS controller and theZhang et al. [5] model.
Fig. no |
Number of robots |
Path traced by the robot using Zhang et al. [5] model in ‘cm’ |
Path traced by the robot usingANFIS controller in ‘cm’ |
Fig.12 (i), (ii) |
1 |
91.3 |
78.9 |
Table 5: The Resultsof path traced by the robot using the ANFIS controller and theMontaner et al. [6] model.
Fig. no |
Number of robot |
Path traced by the robot using Montaner et al. [6] model in 'cm' |
Path traced by the robot using ANFIS controller in 'cm' |
Fig.13 (i), (ii) |
1 |
100.7 |
95.2 |
Fig.14 (i), (ii) |
1 |
101.5 |
97.8 |
Table 6: Experimental mobile robot specifications.
Name |
Specification |
Microcontroller |
Arduino UNO ATmega328 |
Flash Memory |
32 KB (ATmega328) |
Operating Voltage |
5V |
SRAM |
2 KB (ATmega328) |
InputVoltage (Recommended) |
7-12V |
Input Voltage (Limits) |
6-20V |
Digital Input Pins |
14 (of Which 6 Provide PWM Output) |
Analog Input Pins |
6 |
Motors |
2 DC, 30RPM Centre Shaft Economy Series DC Motor |
Motors Driver |
L298, Up to 46V, 2A Dual DC Motor Driver |
Motor Speed |
Max: 30RPM, Min: 12RPM |
Wheel |
Wheel Diameter: 106mm, Wheel Thickness: 44mm, Hole Diameter: 8mm |
Sensors |
1 IR Range Sensor Distance Measuring Range: 20cm to 150cm |
Sensors |
2 Ultrasonic Range Finder Sensor Distance Measuring Range: 2cm to 400cm |
Bread Board |
Small Size Bread Board |
Communication |
USB connection Serial Port |
Size |
Height: 7.5cm, Length: 27cm, Width:33cm, |
Weight |
Approx. 1.4kg |
Payload |
Approx. 400g |
Power |
Rechargeable Lithium Polymer 3 Cell, 11.1V, 2000mAh, 20C Battery |
Experimental Results and Discussion
The proposed ANFIS controller has implemented on the real
experimental mobile robot Figure 15. The robot has two front
driving wheels attached to the separate motors used for Steering
Angle (SA) and driving (motor speed) control and single caster
wheel for balance the robot. Two separate DC motors are used to
drive each front wheel to facilitate turns, backward and forward
movement. The speed of the motors is controlled by the dual
DC motor driver L298, which is connected to the Arduino UNO
(ATmega328) microcontroller. If there are no obstacles detected
nearby the robot, then the right and left motors will move at the
same speed; otherwise, it will slow down. If there are obstacles,
detect right side, then the robot turns positive steering angle
means the right and left motors will move high and low speed
respectively. If there are obstacles, detect left side, then the
robot turns negative steering angle means right and left motors
will move low and high speed respectively. The mobile robot is
equipped with a single sharp Infrared (IR) Range sensor for close
obstacle detection, and two ultrasonic range finder sensors are
used for the purpose of sensing obstacles around the robot. The
sensors are evenly mounted on the front, right and left side of
the robot. Ultrasonic range finder sensors are equipped on the
left, and right side of the mobile robot, which measures distance
between 2cm to 400cm and single sharp Infrared (IR) Range
sensor are attached on the front of the mobile robot, which reads
the distance between 20cm to 150cm. The other specifications of
the experimental mobile robot are given in Table 6.
To demonstrate the effectiveness of an ANFIS network
controller a variety of experiments have been done with the real
experimental mobile robot Figure 15. The resulting algorithm
has been burnt on the C/C++ running microcontroller based
experimental mobile robot and tested in the various cluttered
environments for the comparison of the performances. The
experimental verification of the above simulation results has been
shown in Figure 16 and Figure 17. In Figure 16, the single robot
has been used for the navigation. Similarly, in Figure 17, multiple
robots have been used for the navigation. The dimension of the
environments is 300cm width and 300cm height. We have set a
threshold distance between the robots and obstacles, if the robot
detects the obstacle in the threshold range, then the proposed
ANFIS controller is active, and the robot moves/turns according
to the ANFIS controller output (steering angle). From the above
simulation and experimental results, it can be clearly seen that the
ANFIS network controller efficiently drives the robot safely in the
different cluttered environment. The navigation path length result
between the simulation and experimental is listed in Table 3.

Figure 16: Experimental results for the navigation of mobile robot
same as a simulation environment shown in Figure 8 (i).
Figure 17: Experimental results for the navigation of two mobile robots
same as a simulation environment shown in Figure 9.
Conclusion
In the current study, we have described the application of the
ANFIS network controller for multiple mobile robots navigation
and the obstacle avoidance in the cluttered environments. The
ANFIS controller has three inputs and a single output. The controller receives inputs (obstacle distances) from different
sensors to provide appropriate Steering Angle (SA) control
command as output. Under the supervision of the proposed
ANFIS controller, the mobile robots are autonomously reaching
the goals with optimal and smooth travelling path. The eight
rules and different behaviors such as obstacle avoidance, robots
inter-collision avoidance, and the goal seeking behaviour has
been used for the navigation of multiple mobile robots, which is
less than the other conventional approaches i.e. forty-seven rules
made by Zhang et al. [5] and forty-five rules made by Montaner,
et al [6]. In addition, the effectiveness and efficiency of the
ANFIS network controller have been verified through various
simulations and experiments in the real environment. In future
work, this ANFIS controller may be implemented in the dynamic
environment with the hybridization of other nature-inspired
algorithms for multiple dynamic obstacles and dynamic goal path
planning problems.
- Mucsi K, Ata MK, Ahmadi M. An Adaptive Neuro-Fuzzy Inference System for Estimating the Number of Vehicles for Queue Management at Signalized Intersections. Transportation Research. 2011;19(6):1033–1047.
- Hamidian D, Seyedpoor SM. Shape Optimal Design of Arch Dams Using an Adaptive Neuro-Fuzzy Inference System and Improved Particle Swarm Optimization. Applied Mathematical Modelling. 2010;34(6):1574–1585.
- Jaradat MA, Mohammad R, Quadan L. Reinforcement Based Mobile Robot Navigation in Dynamic Environment. ELSEVIER Robotics and Computer-Integrated Manufacturing. 2011;27(1):135–149.
- Matlab 2012 a. ANFIS Toolbox.
- Zhang N, Beetner D, Wunsch DC, Hemmelman B, Hasan A. An Embedded Real-Time Neuro-Fuzzy Controller for Mobile Robot Navigation. IEEE International Conference on Fuzzy Systems. 2005;pp.319–324.
- Montaner MB, Ramirez SA. Fuzzy Knowledge-Based Controller Design for Autonomous Robot Navigation. Expert Systems with Applications. 1998;4(1):179–186.
- Imen M, Mohammad M, Shoorehdeli MA. Tracking Control of Mobile Robot Using ANFIS. IEEE International Conference on Mechatronics and Automation. 2011;pp.422–427.
- Pothal JK, Parhi DR. Navigation of Multiple Mobile Robots in a Highly Clutter Terrains using Adaptive Neuro-Fuzzy Inference System. Robotics and Autonomous Systems. 2015;72:pp.48-58.
- Hu H, Brady M. A Parallel Processing Architecture for Sensor-Based Control of Intelligent Mobile Robots. Robotics and Autonomous Systems. 1996;17(4):235–257.
- Contreras-Cruz MA, Ayala-Ramirez V, Hernandez-Belmonte UH. Mobile Robot Path Planning Using Artificial Bee Colony and Evolutionary Programming. Applied Soft Computing. 2015;30:319-328.
- Parhi DR, Mohanta JC. Navigational Control of Several Mobile Robotic Agents Using Petri-Potential-Fuzzy Hybrid Controller. Applied Soft Computing. 2011;11(4):3546–3557.
- Mohanta JC, Parhi DR, Patel SK. Path Planning Strategy for Autonomous Mobile Robot Navigation Using Petri-GA Optimisation. Computers and Electrical Engineering. 2011;37(6):1058–1070.
- Pradhan SK, Parhi DR, Panda AK. Fuzzy Logic Techniques for Navigation of Several Mobile Robots. Applied Soft Computing. 2009;9(1):290–304.
- Hu E, Yang SX, Chiu DK, Smith WR. Real-Time Tracking Control with obstacle avoidance of multiple mobile Robots. IEEE Proceedings of the International Symposium on Intelligent Control. 2002;pp.87-92.
- Boubertakh H, Tadjine M, Glorennec P, Labiod S. A Simple Goal Seeking Navigation Method for a Mobile Robot using Human Sense, Fuzzy Logic and Reinforcement Learning. Journal of Automatic Control, Journal of Automatic Control. 2008;18(1):23-27.
- Kubota N, Morioka T, Kojima F, Fukuda T. Learning of Mobile Robots Using Perception-Based Genetic Algorithm. Measurement. 2001;29(3):237–248.
- Hoy M, Matveev AS, Savkin AV. Collision Free Cooperative Navigation of Multiple Wheeled Robots in Unknown Cluttered Environments. Robotics and Autonomous Systems. 2012;60(10):1253–1266.
- Begum M, Mann GK, Gosine RG. Integrated Fuzzy Logic and Genetic Algorithmic Approach for Simultaneous Localization and Mapping of Mobile Robots. Applied Soft Computing. 2008;8(1):150–165.
- Korayem M. H, Nazemizadeh M, Nohooji HR. Optimal Point-to-Point Motion Planning of Non-Holonomic Mobile Robots in the Presence of Multiple Obstacles. Journal of the Brazilian Society of Mechanical Sciences and Engineering2014;36(1):212–232.
- Ahmadzadeh S, Ghanavati M. Navigation of Mobile Robot Using the PSO Particle Swarm Optimization. Journal of Academic and Applied Studies(JAAS). 2012;2(1):32–38.