Mobile Robot Navigation Using Fuzzy Logic

Abstract This paper studies mobile robot navigation in an unknown environment. The navigation algorithm represents the actions of a human driver including: goal seeking, obstacle avoidance, U-turn, and getting rid of deadlocks. In this paper, the action taken by the robot at any time depends on the instantaneous mapping of the ultrasonic sensory information. Except the U-turn action, all other actions are taken using a fuzzy inference system or fuzzy inference system and reinforcement learning. The getting rid of deadlock action prevents the robot from being in some of the situations in which it will be deadlocked within a certain area. In addition, reinforcement learning is used to ease the construction of some of the fuzzy rules by on-line modification of the rule-bases. Simulation results prove the effectiveness of the proposed approach.