Visual-based planning and control for nonholonomic mobile robots

Abstract

An integrated visual-based approach to motion planning and control of a nonholonomic wheeled mobile robot is presented. Vision is used to localize the obstacles and the robot and for real-time feedback control of the vehicle motion. An A ∗ based motion planner is used to search the discretized robot configuration space for collisionfree paths. Nonholonomic constraints are taken into account by properly defining cell adjacency in the discretized configuration space and in the choice of the heuristic functions. Path planning results for unicycle and car-like kinematics are presented and compared in simulated environments. The whole system is tested with the mobile robot SuperMARIO in an unknown indoor environment. In the experiments, tracking of the planned trajectories is achieved via visual feedback from a fixed camera and using a dynamic linearizing control law.