Optimal Control-Based Highway Pilot Motion Planner With Stochastic Traffic Consideration

Abstract

This research proposes an optimal control-based motion planner with consideration of the stochasticity of surrounding human-driven vehicles (HVs). The proposed motion planner is designed for the highway piloting of automated vehicles (AVs). It overcomes the shortcomings of conventional methods and is able to 1) plan motions with behavioral decisions, 2) improve safety by considering the stochasticity of surrounding HVs, and 3) improve control accuracy through vehicle dynamics consideration. The stochasticity of surrounding HVs is incorporated into a chance collision avoidance constraint, which is successfully linearized. This design greatly improves the computing efficiency of the proposed method. The proposed motion planner is evaluated by field tests and a simulation in the context of traffic. The results demonstrate that 1) the proposed motion planner reduces 70% of the magnitude of risk and 88% of the duration of risk; 2) to gain the aforementioned safety improvement, the proposed motion planner sacrifices only 1.4% of mobility; 3) the control error is less than 12 cm; and 4) the computation time is about 5 ms. The results indicate that the proposed motion planner is ready for real-time implementation.