Prediction of road adhesion coefficient for vehicle ABS control using an MFO-Elman neural network

Precisely estimating road adhesion coefficients is essential for enhancing the control of wheel slip ratios. However, traditional neural networks exhibit instability in weight vector updates, resulting in suboptimal estimations that are inadequate for real-time traffic control. In this study, a novel method for predicting road adhesion coefficients utilizing the MFO-Elman neural network is devised to enhance the prediction accuracy. First, the neural network model is constructed based on a vehicle dynamic model. Subsequently, the initial moths are introduced to increase the diversity of the initial population, accelerate the convergence speed, and reduce the number of calculations. Combining the moth flight update mechanism can effectively mitigate the risk of the algorithm converging to a local optimum. Then, the Fourier transform method is employed to estimate the correlation between the optimal slip ratio and the tire-road friction coefficient. Both simulation and experimental findings corroborate the reliability of the newly devised MFO-Elman neural network method for predicting road adhesion coefficients, which yields a higher estimation precision for the road adhesion coefficient.