Adaptive neuro-fuzzy inference system based robotic navigation

Abstract

The oldest challenge in mobile robotics is the ability of robot to navigate autonomously in a dynamic environment. This paper, discusses about navigation of mobile robot using Adaptive Network-Based Fuzzy Inference System (ANFIS) which is basically fuzzy inference system implemented in framework of adaptive networks. Hybridization of Fuzzy Logic and Artificial Neural Network engenders the autonomous robot to give a human-like reasoning to problems and acquire implicit knowledge using stipulated input-output pairs. A non-holonomic robot consisting of Sonar and Magnetometer sensors verifies feasibility of developed code. The front obstacle distance from Sonar and steering angle from Magnetometer provide input to the Fuzzy Layer. The weights of the adaptive nodes are tuned by one-pass Least Square Estimator followed by iterative Steepest Descent approach. The autonomous robot is able to avoid obstacles and reach the target location from starting point using the adaptive parameters obtained from simulation.