Novel Double Adaptive Algorithm for Tire–Road Friction Estimation

Abstract

Numerous studies on traffic safety have demonstrated the likelihood of accidents closely associated with the tire–road friction coefficient (TRFC). However, under complex working conditions, it is difficult to accurately obtain the value of TRFC. In this paper, a novel double adaptive algorithm based on an adaptive strong-tracking cubature Kalman filter (ASTCKF) and an adaptive backpropagation neural network (ABPNN) is proposed to estimate the TRFC. First, a nonlinear three-degree-of-freedom vehicle model, complemented by a magic formula (MF) tire model, is established. Then, the strong tracking theory (STT) and the adaptive noise matrices method are incorporated into the cubature Kalman filter to form the ASTCKF to estimate the vehicle driving states. Then, the BP neural network combined with an adaptive learning rate is designed to estimate the TRFC. Finally, the proposed algorithm is verified through Carsim/Simulink. The cosimulation results show that the TRFC estimation algorithm based on ASTCKF and ABPNN has remarkable estimation accuracy and is suitable for different complex road conditions.