Real time dynamic adhesion coefficient estimation and BP neural network optimized lateral stability control for distributed drive electric vehicles

To address the instability issues of distributed-drive electric vehicles (DDEV) operating on roads with abrupt changes in adhesion coefficients, a lateral stability control strategy and torque distribution method based on backpropagation (BP) neural network optimization were proposed. First, an Unscented Kalman Filter (UKF) estimation algorithm incorporating real-time variation detection of adhesion coefficients was developed. To ensure rapid response and accurate estimation of current adhesion coefficients during sudden road condition changes, threshold-based real-time detection of adhesion coefficient fluctuations was introduced. Second, a hierarchical stability control strategy specifically designed for varying adhesion coefficient conditions was established. The upper-layer controller employs a Bat Algorithm (BA) optimized BP neural network, which takes the sideslip angle and yaw rate as control targets to calculate the required yaw moment for vehicle stabilization, thereby enhancing real-time computational efficiency and solution accuracy. The lower-layer controller utilizes the estimated road adhesion coefficients to implement a quadratic programming algorithm, optimizing wheel torque distribution to minimizing tire load rate. Finally, a co-simulation platform was constructed using Carsim/Simulink for validation. The results demonstrate that the proposed estimation algorithm can precisely estimate road adhesion coefficients under extreme conditions of abrupt coefficient changes. The developed stability controller significantly enhances both handling stability and driving stability of DDEV.