Risk-informed decision-making and control strategies for autonomous vehicles in emergency situations

This paper proposes risk-informed decision-making and control methods for autonomous vehicles (AVs) under severe driving conditions, where many vehicle interactions occur on slippery roads. We assume that the AV should approach a specific safe zone in case of vehicle malfunctioning. In normal situations, the driving behavior of the AV is based on the deterministic finite-state machine (FSM) that makes an appropriate real-time decision depending on the driving condition, which is efficient when compared with some conventional decision-making algorithms; alternatively, in emergency situations, the AV first has to be prevented the rear-end conflicts while the specific safe zone is simultaneously determined by evaluating the level of risk via two safety level indicators, i.e., Time-To-Collision (TTC) and Deceleration Rate to Avoid the Crash (DRAC). The safe path that guides the AV to avoid car crashes is generated based on the trajectory optimization theory (i.e., Pontryagin maximum principle), and the AV follows it based on the linear time-varying model predictive control (LTV-MPC), which ensures the AV's lateral stability. We verify the effectiveness of the proposed decision-making and control strategies in various test scenarios, and the results show that the AV behaves appropriately according to the behaviors of surrounding vehicles and road condition.