Adaptive Potential Field with Collision Avoidance for Connected Autonomous Vehicles

Potential field (PF), as a risk assessment method, is proposed to enhance autonomous vehicles' (AVs) safety in collision avoidance. However, current PF targets mainly standalone-mode AVs (SAVs) by evaluating their relative position and velocity. In addition, the risk energy of the PF is usually assigned an infinite value along the z-axis. Therefore, this study presents an adaptive potential field (APF) for connected autonomous vehicles (CAVs). Valuable information (heading angle, steering wheel angle, etc.) other than relative position and velocity is supplemented to PF. Furthermore, we separate the APF from the cost function of the model predictive controller (MPC) to compute the desired reference signals directly, saving more computation time. The proposed APF-MPC is co-simulated in a comparative driving scenario via MATLAB/Simulink and CarSim simulator compared with the latest PF-MPC method.