A Convex Trajectory Planning Method for Autonomous Vehicles Considering Kinematic Feasibility and Bi-state Obstacles Avoidance Effectiveness

The Trajectory Planning Module is a key component of the Autonomous Driving System which is central to the realization and widespread use of Autonomous Vehicles. However, in the models adopted by mainstream trajectory planners, only crude representations of the geometrical sizes of vehicles are usually made, a limitation that cannot guarantee the kinematic feasibility of the driving process. But accounting for kinematic constraints will present a further challenge as the model itself will be non-convex due to the inherent angularity of vehicles. To cope with this challenge, a trajectory planning method considering kinematic feasibility based on the convex opntimization framework is proposed. Instead of employing solving techniques that loosen non-convex effects, our method directly addresses the problem of convex kinematic feasibility by constructing a collision-free constraint regarding a vehicle's precise size and a variable boundary constraint regarding the vehicle's inherent limitation. Moreover, we present a "Bi-state Obstacles Avoidance" strategy designed to overcome the disadvantages in typical trajectory planners that only consider a static obstacle avoidance constraint in the path generation phase of trajectory planning. By introducing an external parameter "time", the staticization representation of dynamic obstacles can effectively be implemented, which greatly improves the ability to regulate velocity. The feasibility and effectiveness of the proposed new planner were tested using both simulated and real-world driving data, the results of which show that the proposed method can plan safe and efficient trajectories in real-time, an essential requirement for implementation in real-world autonomous vehicles.