A fuzzy logic approach to command arbitration for an autonomous mobile robot

Abstract

One of the ultimate goals in the field of robotics is to create autonomous robots. Such robots would be able to accept high level input commands from either a human operator or a centralized control system and perform that task with no outside influence. Ideally, the input commands would only specify what was needed to be done, not how. Developing an autonomous mobile robot requires studying and intergrating the concepts of autonomous sensing, reasoning, and control. One of the basic issues in autonomous reasoning, which has ramifications also in sensing and control, is that of motion planning. Mobile robot motion planning is the process of determining a set of steps for the robot to execute in order to achieve its goal position. In one of its simplest forms, motion planning involves computing a set of trajectories and speeds for the robot to take in order for it to arrive at its predetermined destination. There are many methods to accomplish this form of motion planning, but all have some drawbacks which could limit the capabilities of the robot. The major goal of this research has been to extend an approach developed by D. Payton and J. Rosenblatt which combines some of the best elements of previous approaches. Their research was in a behavioral approach which attempts to minimize information loss in command formulation by using a command arbitration network, termed command fusion. We have extended Payton-Rosenblatt's approach using fuzzy logic. The benefits of our approach include simplicity, extensibility and understandability in the command formulation process. Another contribution of this research has been to identify the limitations of existing defuzzification techniques for dealing with prohibitive information. We have developed a new defuzzification strategy that has been able to remedy this problem. Keywords:Fuzzy Logic, Fuzzy Control., Behaviors, Autonomous Mobile Robots