Fuzzy Logic Based Reactive Control of a Mobile Robot for Navigation in an Unstructured Environment

Abstract— This paper presents a proposed fuzzy logic based control algorithm as a behavior based control scheme for real time navigation of a mobile robot in an unstructured environment. Behavior based control makes use of a set of distributed, interacting modules called behaviors that collectively help achieve the desired response. Each behavior receives inputs from sensors and/or other behaviors in the system and provides outputs to robot’s actuators or to other behaviors. For reactive navigation control, rule based methods can be employed in order to avoid lengthy computations and complexity associated with the involvement of world-models. To verify the functionality and effectiveness of multiple behaviors that is Left Wall Following, Right Wall Following and Corridor Following, simulations have been performed on a 2D configurable simulator. Workspace for each of these simulations especially corridor navigation has been designed so as to incorporate all different types of situations that the robot is expected to face in a real world navigation problem. The method has been proposed for implementation on an autonomous differential wheeled mobile robot “UBot". It is a circular shaped, compact structure, designed and developed for experimentation with the newly emerging intelligent navigation algorithms.

Keywords—-fuzzy logic based control, behavioral control scheme, unstructured environment, navigation

1. Introduction

Mobile robots are a major focus of ongoing research in robotics. The rapidly developing field of mobile robotics is gaining recognition by taking the jobs of human beings working in unfavorable conditions like those found in nuclear installations, minesweeping ¹ and bomb disposal ² etc. Mobile robot assistance is being employed in industry, military and security environments ³ ; they are being used in space ⁴ and underwater exploration ⁵ and have proved to be of great help in gathering useful information that helped carry out further research in various other fields. Recent work in obstacle avoidance and navigation in 3D workspaces has opened up a completely new world of robot applications. Such robots can be used, although with few limitations, in unstructured environments such as those encountered during the search of victims in an earthquake affected, collapsed house ⁶.

For mobile robots, the most fundamental and pressing problem is the navigation of the robot. To accomplish the task of reporting the authorities of an accident in a house for handicapped, the human assistive robot ⁷ would need to locate itself first, map the location of mishap with respect to its location and navigate through the unstructured environment with the aid of intelligent navigation and control algorithms. The successful navigation thus depends upon four vital blocks ⁸:

illustration not visible in this excerpt

Fig. 1: Successful Navigation

To successfully implement each of these vitals as shown in the Fig. 1, one of the two popular controlling approaches as given below is chosen based on the characteristics of the environment the autonomous mobile robot is expected to confront.

1) Deliberative Navigation Control, and
2) Reactive Navigation Control.

In deliberative navigation control, it is assumed that the environment of the robot is fully or partially known. The global map of the environment lias already been fed into the robot's control algorithm, to reach the desired destination requires the robot to preplan the minimum cost path to follow using the developed model of the environment and data acquisition from sensors. This approach may work in structured and small environments but is highly unfavorable in larger and dynamic environments. In an unstructured environment, where perceiving the environment is one of the key challenges faced by the robot and the robot has to continuously monitor its effect on the environment while acquire information necessary to support decision making and reacting to undesired events ⁹, it is better off by making decisions employing reactive control approach.

Reactive navigation control uses contemporary data from sensors as input and uses that data to make decisions. Only the nearby features are considered and the robot makes decisions related to navigation based on its position with respect to these local features. One of the advantages of this approach is that the robot does not need to be loaded with any prior knowledge of the environment. Since the decision depends upon the current data only, lesser computation is required and quick decisions can be made.

Amongst a number of intelligent navigation approaches being adopted these days such as reinforcement learning, fuzzy logic, neural networks, neuro-fuzzy control, fuzzy logic is gaining popularity among researchers due to its simplicity, capability to perform a wide variety of tasks without explicit computations and measurement ¹⁰ and robustness ¹¹. Without accurate model of the environment, dealing with the high degree of complexity and non-linearity associated with the real-world navigation would require the expertise of a human reasoning/decision making and fuzzy control could only provide those means to capture the human mind’s expertise. Ever since the researchers have recognized the potential the fuzzy logic based control holds, many approaches such as those presented in ^{12 13 14}, have been put forward to solve mobile robot navigation problems such as target tracking, wall tracking, target steering and obstacle avoidance.

This paper proposes a fuzzy logic based control algorithm as a behavior based control scheme for navigation of a mobile robot (Fig. 8) in an unstructured environment. The paper is organized as follows: 1st section covers introduction, design of the fuzzy controller is presented in the 2nd section, 3rd section covers the overview of the mobile platform with the overview of simulation software used. Simulation results have been presented in the 4th section while 5th section presents the conclusions and future work.

2. Design of the fuzzy controller

Fuzzy systems deal with human reasoning in the form of sets of if-then rules. Each rule activates a response when a certain condition is satisfied. It is therefore very important for the control system designer to understand the possible situations and be aware of complexities involved before he moves on to design of the controller. Fuzzy controller design steps could be summed up as shown in the Fig. 2.

illustration not visible in this excerpt

once the crisp input data has been received and membership functions have been defined, fuzzification of variables is equivalent of converting the crisp input values into a form that the inference engine can understand. Inference engine is used to evaluate the rules and combine their result in order to map from a real world point to fuzzy set. Defuzzification is the process of converting fuzzy rules into actual outputs.

Same procedure has been followed and presented for the design of the fuzzy logic based controller in this paper. Multiple behaviors have been implemented in initially separate and then a single simulation. Behavior based control scheme is designed so that the right behavior building block could be selected for a particular system. The behaviors considered for the behavior based control of the autonomous mobile robot "UBot”. in an unstructured environment, arc described in tltc following subsections.

2.1 Wall Following Behavior

For a mobile robot designed primarily for navigation in an unstructured environment, at times, it becomes vital for a robot to follow a particular wall in a maze until an event forces it to change the behavior. One such event is the appearance of an obstacle. In any case, wall following behavior building block is one of the essential blocks and cannot be ignored.

2.1.1. Membership Functions for Wall Following

For both left wall following and right wall following, the basic algorithm is same. Distance of the robot with respect to the wall at any particular point in time is obtained using ultrasonic range finders mounted on the sides of the mobile platform. Two variables named as “s” and "ds" arc defined where "s” stores the current reading of the sensor or the current distance from the wall while "ds" stores the difference between the current distance value and previous distance value. Hence, “s” is the measure of the position of the robot with respect to the wall and "ds" represents the direction of motion of tltc robot and gives an idea of the speed of the robot in that direction. Tltc negative value of "ds" means motion is towards the wall and positive value represents motion away from the w all. Membership functions for both "ds” and "s” are presented in Fig. 3 and Fig. 4 respectively.

illustration not visible in this excerpt

Fig. 3: Membership Functions for "ds"

illustration not visible in this excerpt

Fig. 4: Membership Functions for "s"

Values obtained using ultrasonic range finders on the left of the mobile platform are used for activation/deactivation of left- wall-following control while values obtained using range finders on the right are used for right-wall-following control. The output membership functions i.e. velocity membership functions for left wheel and right wheel are as shown in Fig. 5 and Fig. 6 respectively; left wheel velocity is stored in a variable “lw vel” and the right wheel velocity is stored in “rvv vel”

illustration not visible in this excerpt

Fig. 5: Membership Functions for “lw vel”

illustration not visible in this excerpt

Fig. 6: Membership Functions for “rw vel”

2.1.2. Rule Base

a) Left-Wall-Following Rules:

Rules derived from basic human judgment are used to create the rule base for each of the behaviors implemented. Left wall following behavior realization is achieved using following rules:

If (s = near and ds = neg_zero)

then (lw_vel = fast and rw_vel = slow)

If (s = medium and ds = pos_large)

then (lw_vel = slow and rw_vel = v_fast)

If (s = far and ds = pos_small)

then (lw_vel = slow and rw_vel = v_fast)

The rule base created for left-wall-following is made up of total 15 such rules, 3 are presented here.

b) Right-Wall-Following Rules:

Similar to the left-wall-following rules, rule base made up of total 15 if-then rules is created for right-wall-following, 3 of the rules are given below.

If (s = near and ds = pos_small)

then (lw_vel = medium and rw_vel = fast)

If (s = medium and ds = neg_large)

then (lw_vel = slow and rw_vel = v_fast)

If (s = far and ds = zero)

then (lw_vel = fast and rw_vel = slow)

2.2. Corridor Following Behavior

In situations where a wall exists on both the right side and left side of the robot, corridor following behavior building block would be the most appropriate choice. In wall following, control is required to make sure that the robot stays at the safe distance from the wall, while progressing forward. But when there is a wall on both sides of the robot, it is much safer to keep same distance from both the walls in order to avoid collision with either side.

2.2.1. Membership Functions for Corridor Following

The corridor-following algorithm depends on the difference between the values obtained using range finders on the right and the values obtained using range finders on the left of the robot: the difference obtained is stored in the variable "diff”. For the robot to follow ideal corridor-following, “diff” would have to be kept equal to zero. Membership functions for the variable "diff’ arc illustrated in the Fig. 7.

illustration not visible in this excerpt

Fig. 7: Membership Functions for "diff'

2.2.2. Rule Base

Based on the membership functions for the variable ‘diff' (Fig. 7) and using the same variables lw_vel and rw_vel to store the output velocities of left wheel and right wheel respectively, following rule base has been created for corridor­following.

If (diff = neg_large)

then (lw_vel = v_fast and rw_vel = slow)

If (diff = neg_small)

then (lw_vel = fast and rw_vel = slow)

If (diff = zero)

then (lw_vel = medium and rw_vel = medium)

If (diff = pos_small)

then (lw_vel = slow and rw_vel = fast)

If (diff = pos_large)

then (lw_vel = slow and rw_vel = v_fast)

2.3. Defuzzification

Once the rules have been used to form a solution set and inference mechanism has done its job, defuzzification interface is required to convert the solution set into a crisp output. Among the various methods being employed for the defuzzification, the Center-of-Gravity or Centroid method is the most popular method. Mathematical interpretation of COG method is as given below.

illustration not visible in this excerpt

This allows for the incorporation of multiple membership functions active at the same time to have their effect on the outputs. For example, in case of corridor-following behavior, the neg_small and zero membership functions may simultaneously get activated to some extent. The extent to which each of the active membership functions will affect the output velocities is dictated by the weights of each of them. These weights are then used to calculate the active area of the membership function i.e. ”μ(γΟ in the equation. This weight assigning mechanism helps combine the active membership functions in the defuzzification process.

3. Overview of the Mobile Platform and the Simulation Software

3.1. Mobile Platform

The mobile platform "UBot" (Fig. 8) that is designed and developed by the author, as an educational robotic kit, for research in problems related to intelligent navigation, is a circular and compact, differential drive mobile robot with a detachable ball potting mechanism. All the required electronics for the implementation of the proposed fuzzy control algorithm and other such intelligent algorithms, has also been designed, developed and mounted on the platform. The electronics developed has been tested with basic motion control algorithms to make sure that the students/professionals that plan to pursue this work to implementation of the proposed algorithm do not face any problem due to faulty hardware.

illustration not visible in this excerpt

Fig. 8: Mobile Robot "UBot"

3.2. Simulation Software

Simulation software used for the verification of the functionality of the proposed and designed (Section 2) fuzzy logic based control, is a 2D configurable software "MobotSim" of MobotSoft Company. It has been chosen over other simulation software because of its usability and relevance when it comes to circular, differential drive mobile robots. It has got a libran' of functions for programming of robots in "Sax Basic" language and a library of different types of objects that could very well be used to emulate in 2D, a lab environment, an office, a long corridor or a round corridor. In Fig. 9, the figure on the left shows an arrangement of 6 ultrasonic range finders on the robot while the figure on the right shows how the sonar radiation cone has fallen in size due to interaction with corridor walls.

illustration not visible in this excerpt

Fig. 9: MobotSim Environment

4. Simulation results

Simulations have been carried for a variety of environments to test the proposed control algorithms. For right-wall­following and left-wall following, gradual turns and straight walls are the basic tests performed while for corridor­following, a complex workspace was designed. For each of different types of corridor environments such as, a corridor with walls inclined at an angle Θ, a corridor with a sudden change in the width, a corridor with right angled trajectory and, a circular corridor, the algorithm worked as fine as it worked for the one with straight side walls. All these environments have been selected from [14].

4.1. Wall Following Behavior

As shown in the Fig. 10, the mobile robot reaches the end of the wall and neither hits any part of the wall nor diverts from its course on its way there. Despite the sharp edges in Fig. 10(b), the algorithm designed worked as well as it worked for gradual turn in Fig. 10(a).

illustration not visible in this excerpt

Fig. 10(a): Left Wall Following Simulation

illustration not visible in this excerpt

Fig. 10(b): Right Wall Following Simulation

4.2. Corridor Following

To test the working of corridor following behavior building block, all the different situations proposed in [14] were incorporated into a single environment. The line that seems to be the tail of the robot in the Fig. 11 actually shows the path followed by it. It could be seen that despite sudden changes in the environment, the algorithm worked well enough to allow the robot to reach the end without hitting any of the side walls.

illustration not visible in this excerpt

Fig. 11: Corridor Following Simulation

5. Conclusions and Future Work

In this paper, a fuzzy logic based reactive control algorithm has been proposed as a behavior based control scheme for navigation of the mobile robot ”UBot” in an unstructured environment. The functionality and effectiveness of the control system designed is verified by simulation in a 2D simulator "MobotSiin". The design of the control system and the simulation results obtained for some of the behaviors like left wall following, right wall following and corridor following have been presented. The same procedure could be followed and system designed could be improved by adding more behavior building blocks in the algorithm designed, thus preparing it for navigation in more complex environments. Since all the hardware required for the implementation of the fuzzy control system proposed in this paper has been designed, fabricated, and mounted on the robot, it is time for the implementation of the algorithm.

REFERENCES

[...]

---

¹ R. Abbaspour, "Design and Implementation of Multi-Sensor based Autonomous Minesweeping Robot," in 2010 International Congress on Ultra Modern Telecommunications and Control Systems and Workshops (ICUMT), Moscow, Oct. 2010.

² C. B. R. M. a. J. B. Brian Day, "A Depth Sensing Display for Bomb Disposal Robots," in IEEE International Workshop on Safety, Security and Rescue Robotics, SSRR,, Sendai, 2008.

³ R. Schultz, R. Nakajima and J. Nomura, "Telepresence Mobile Robot for Security Applications," in 1991 International Conference on Industrial Electronics, Control and Instrumentation, IECON '91, Kobe, 28 Oct-1 Nov 1991.

⁴ A. Elfes, J. M. Dolan, G. Podnar, S. Mau and M. and Bergerman, "Safe and Efficient Robotic Space Exploration with Tele- Supervised Autonomous Robots," in Proceedings of the AAAI Spring Symposium, Pola Alto, California, March, 2006.

⁵ C. Kunz, C. Murphy, R. Camilli, H. Singh, J. Bailey, R. Eustice, M. Jakuba, K. Nakamura, C. Roman, T. Sato, R. Sohn and C. Willis, "Deep Sea Underwater Robotic Exploration in the Ice- covered Arctic Ocean with AUVs," in IEEE/RSJ International Conference on Intelligent Robots and Systems, 2008. IROS 2008. , Nice, Sept. 2008.

⁶ D. Lee, "Microsoft Kinect-Powered Robot to Aid Earthquake Rescue," BBC News Technology, 25 February 2011. [Online]. Available: http://www.bbc.co.uk/news/technology-12559231.

⁷ T. Chen, M. Ciocarlie, S. Cousins, P. Grice, K Hawkins, K. Hsiao, C. Kemp, C. King, D. Lazewatsky, A. Leeper, H. Nguyen, A. Paepcke, C. Pantofaru, W. Smart and L. Takayama, "Robots for Humanity: User-centered Design for Assistive Mobile Manipulation," in IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Vilamoura , Oct. 2012 .

⁸ Siegwart, Roland and I. R. Nourbakhsh, Introduction to Autonomous Mobile Robots, The MIT Press, 2011.

⁹ D. Katz, J. Kenney and O. Brock, "How Can Robots Succeed in Unstructured Environments?," in Workshop on Robot Manipulation: Intelligence in Human Environments at Robotics: Science and Systems, June 2008, Zurich, 2008.

¹⁰ D. N. a. B. K Tang Sai Hong, "Application of Fuzzy Logic in Mobile Robot Navigation," in Application of Fuzzy Logic in Mobile Robot Navigation, Fuzzy Logic - Controls, Concepts, Theories and Applications, Prof. Elmer Dadios (Ed.), Universiti Putra Malaysia, Malaysia, InTech, 2012.

¹¹ C. Coleman and D. Godbole, "A Comparison of Robustness: Fuzzy Logic, PID, and Sliding Mode Control," in Third IEEE Conference on Fuzzy Systems, 1994. IEEE World Congress on Computational Intelligence., , Orlando, FL , 1994.

¹² W. Li, "Fuzzy Logic Based Robot Navigation In Uncertain Environments by Multisensor Integration," in Proceedings of the 1994 IEEE International Conference on Multisensor Fusion and Integration for Intelligent Systems (MFI '94), Las Vegas, NV, 1994.

¹³ D. R. Parhi, "Navigation of Mobile Robots Using a Fuzzy logic Controller," Journal of Intelligent and Robotic Systems , vol. 42, no. 3, pp. 253-273, 2005.

¹⁴ U. Farooq, G. Abbas, S. O. Saleh and M. U. Asad, "Corridor Navigation with Fuzzy Logic Control for Sonar Based Mobile Robots," Proc. 7th IEEE Conference on Industrial Electronics and Applications, pp. 2087-2093, Singapore, July 2012.