Comparative Analysis of Fuzzy Inference Systems Applications on Mobile Robot Navigation in Unknown Environments

Mobile robot navigation in unknown environments is a central problem in mobile robot control and several studies were carried out in the past decades. This paper compares three mobile robot navigation methods that use fuzzy inference systems (FIS) to develop a controller for the robot: Multiple Fuzzy Inference System (MFIS), Adaptive Neuro-Fuzzy Inference System (ANFIS) and Fuzzy Artificial Potential Field (FAPF). The mobile robot navigation task was divided into two behaviors: go-to-target and obstacle avoidance. Four scenarios were used to validate the methods: a cluttered environment, a narrow path with the robot starting from the center point between walls, a narrow path with the robot starting from the left point misaligned with the center of the two walls, and an environment with an obstacle positioned to generate a point of local minima followed by a narrow gap. Simulation results were presented using MATLAB and CoppeliaSim software and the classical Artificial Potential Field (APF) method was used as the basis for the comparison. The FAPF method used a fuzzy inference system to weigh the value of the attraction force and the repulsion force of the artificial potential fields method ended up presenting the best general performance related to the distance traveled considering all four scenarios, being the method that the robot passes more centralized when between obstacles on both sides, while the MFIS method that used two fuzzy inference systems for each behavior ended up presenting the best performance related to the execution time of the tasks for all four scenarios.