Model Predictive Decision-Making Considering Lane-Changing Time Under Emergency Collision Avoidance for Intelligent Vehicles
Qikun Dai, Jun Liu, Hongyan Guo, Hong Chen, Dongpu Cao
Abstract
When a vehicle faces an imminent collision, it becomes imperative for intelligent vehicles to make emergency collision avoidance decisions in order to mitigate traffic accidents and reduce injuries. To address collision avoidance in emergency scenarios, this study proposes a model predictive decision-making (MPDM) approach that incorporates the consideration of lane-changing time. First, a simplified integrated longitudinal and lateral decision-making model is established, and its accuracy is validated through comparison with real vehicle data. Second, a mixed integer nonlinear MPDM is designed to optimize emergency collision avoidance decisions. Within this framework, the minimum lane-changing time for intelligent vehicles is analytically derived based on vehicle dynamics, taking into account varying speeds and adhesion coefficients. Third, by reducing the dimensionality of the lane-changing time optimization variables, an equivalent suboptimization problem is introduced, which consequently diminishes the computational complexity of solving the optimization problem. Finally, a comparative analysis was performed between the MPDM method and several alternative approaches, employing Simulink-SCANeR cosimulation. Furthermore, the MPDM method was validated on a real vehicle. The results obtained highlight a significant enhancement in the safety and stability of collision avoidance due to the MPDM.