A Real-Time Obstacle-Avoidance Trajectory Planner for On-Road Autonomous Vehicle

A real-time obstacle-avoidance trajectory planner for on-road autonomous vehicle is proposed in this paper. A cubic B spline core is parametric to generate path with continuous curvature as well as taking the extreme curvature limited by steer system into account. By sampling the target sets via offsetting along the reference path, lots path sets are produced with same heading. As embedded with collision checker and path evaluator, a path selector could pick out the best one to planning speed profile for coupling trajectory to track. Finally, according to the change of path curvature, the speed profile scheduler addresses the conflict of curvature and deceleration. Referencing to the ISO3888-2:2011, a collision avoidance scenario was design to validate the planner. The test results of ten cycles test illustrate that the planner has high enough real-time performance as mean plan time less than 100ms with success rate about 100%.