Reactive Obstacle Avoidance Strategy Based on Fuzzy Neural Network and Arc Trajectory

Abstract

To solve the obstacle avoidance of a wheeled mobile robot in the target navigation process, a method based on arc trajectory and fuzzy neural network is proposed. Although some related works consider linear and angular velocities as control inputs, the sample inputs of the training neural network are too subjective and not very easy to select. In this paper, a very clear sample set can be established by CAD software. Then through network training and kinematics model transformation, the linear velocity of two wheels is obtained as the control input. Finally, the advantages of this method are demonstrated by STDR and ROS simulation.