Driving Environment Uncertainty-Aware Motion Planning for Autonomous Vehicles

Abstract Autonomous vehicles require safe motion planning in uncertain environments, which are largely caused by surrounding vehicles. In this paper, a driving environment uncertainty-aware motion planning framework is proposed to lower the risk of position uncertainty of surrounding vehicles with considering the risk of rollover. First, a 4-degree of freedom vehicle dynamics model, and a rollover risk index are introduced. Besides, the uncertainty of surrounding vehicles' position is processed and propagated based on the Extended Kalman Filter method. Then, the uncertainty potential field is established to handle the position uncertainty of autonomous vehicles. In addition, the model predictive controller is designed as the motion planning framework which accounts for the rollover risk, the position uncertainty of the surrounding vehicles, and vehicle dynamic constraints of autonomous vehicles. Furthermore, two edge cases, the cut-in scenario, and merging scenario are designed. Finally, the safety, effectiveness, and real-time performance of the proposed motion planning framework are demonstrated by employing a hardware-in-the-loop experiment bench.