A real-time streamline-based obstacle avoidance system for curvature-constrained nonholonomic mobile robots

This work presents a streamline-based strategy for curvature-constrained nonholonomic robots to safely navigate in real-world partially unknown environments with static cylinder-shaped obstacles whose locations are detected on-line. We propose the use of three primitive curvature-constrained collision-free paths derived from the manipulation and search of streamlines generated from harmonic potential function of a stationary circle. The primitive paths allow emergent turning and reduce the clearance with an obstacle. Assuming enough clearances between adjacent obstacles, the algorithm is extended to avoid multiple cylinder obstacles by pursuing a primitive path based on lateral distance of the robot and the closest obstacle, which can be calculated and updated in real time using pure pursuit tracker. Experiments validate that the obstacle avoidance system allows a two-wheel driving mobile robot to on-line navigate safely along a path without violating curvature constraint in partially unknown cluttered environments.