Adaptive Model Predictive Control of Four-Wheel Drive Intelligent Electric Vehicles Based on Stability Probability Under Extreme Braking Conditions

Under extreme braking conditions, the sharp decrease in vehicle speed causes the probability of vehicle states to change rapidly between stable and unstable, significantly impacting its overall safety and stability. Concurrently, the intricate nonlinear coupling of the vehicle's longitudinal, lateral, and vertical dynamics poses significant challenges in maintaining stability. To address these challenges and enhance vehicle stability comprehensively, this paper proposes an adaptive stability probability-based model predictive control (ASPMPC) algorithm for four-wheel drive intelligent electric vehicles. This algorithm aims to mitigate the safety risks associated with speed changes and redundant control. Initially, a driving simulator is utilized to gather vehicle stability data, which is then categorized into four states-stable, trending stable, trending unstable, and unstable-using the density-based spatial clustering of applications with noise (DBSCAN) algorithm. This categorization forms the basis of a stable dataset. Subsequently, this dataset is used to train a one-dimensional convolutional neural network (1D-CNN), generating a real-time stability probability spectrum. Then, based on model predictive control framework, a correlation function of control objective weight factors is established according to the real-time stability probability of the vehicle. This enables dynamics adjustment of the vehicle's longitudinal, lateral, and vertical stability under extreme braking conditions. Simulations and hardware-in-loop tests have demonstrated that ASPMPC outperforms existing methods, minimizing errors in lateral velocity, yaw rate, and roll angle, thus enhancing maneuverability and safety under extreme braking conditions.