Vision-based smooth obstacle avoidance motion trajectory generation for autonomous mobile robots using Bézier curves

Abstract

This paper proposes an obstacle avoidance trajectory generation method that provides a smooth trajectory in real time. The trajectory is generated from an environmental top-view image, where a fisheye lens is used to capture a wide area at low height. Corners of the obstacles are detected and corrected using the log-polar transform and are used to generate a simple configuration space that reduces the computation time. An optimal path is computed by using the A[Formula: see text] algorithm and replaced by a smooth trajectory generated based on piecewise quintic Bézier curves. Based on the established goal and visual information, a method for generating the first and second derivatives at the start and end points of each Bézier segment is proposed to generate a continuous curvature trajectory. The method is simple and easy to implement and has an average computation time of 1.17s on a PC (CPU: 1.4 GHz) for a workspace containing five to six obstacles. Experimental results verify that the proposed method is effective for real-time motion planning of autonomous mobile robots.