非结构道路场景下轮式无人车辆避障算法

An obstacle avoidance algorithm based on model prediction control is proposed for the obstacle avoidance task of autonomous ground vehicle in unstructured environments. The trajectory of vehicle, steering angle and reference acceleration can be generated by using optimization method without the offline map and the reference trajectory in unstructured environments, and the result satisfies the vehicle motion constraint in passable area. Obstacle avoidance problem is divided into multi-phrases optimal control problem by using the environment perception results according to unstructured road scenarios. The vehicle constraints and cost function for each phrase of the optimal control problem are clarified. The constraints are transformed into the formulation related to vehicle longitudinal and lateral control. Then the optimal control problem is solved by mathematical method. The results of real vehicle tests show that the proposed algorithm can generate trajectory in unstructured environments. It is also shown that the algorithm is capable for obstacle avoidance and keeps vehicle under the limits of constraint.