Linear Time-Varying MPC-Based Autonomous Emergency Steering Control for Collision Avoidance

In this paper, we propose an autonomous emergency steering (AES) control strategy for the automated vehicle to avoid collisions with obstacles while maintaining the vehicle's yaw stability. Specifically, by capturing the surrounding traffic environment (i.e., dynamic/static obstacles and lane-road boundary) using an artificial potential field approach, a safe vehicle path trajectory in the form of a polynomial function was generated via Pontryagin's maximum principle (PMP). The generated path was complemented by a linear time-varying model predictive control (LTV-MPC) scheme that can predict the state behaviors near the operating points during the prediction horizon. Therefore, the proposed approach can effectively and dynamically adapt to the changes in the system model while respecting the state-and-control constraints. Furthermore, we assumed harsh driving environments where the vehicle drives on a slippery road at a relatively high speed, which may cause lateral instability to the vehicle. Hence, the developed controller has to consider both the vehicle's yaw stability and the path-tracking performance. To achieve this goal, we imposed the adaptive constraints in the designed LTV-MPC, depending on the road surface and driving conditions to prioritize the satisfaction of vehicle yaw stability instead of the path-tracking performance. The proposed approach was verified under various conditions using high-fidelity vehicle dynamics and control testing software (i.e., CarSim). Its effectiveness in vehicle yaw stability and path-tracking performance is compared to conventional baseline approaches.