A Data-Driven Method for the Estimation of Truck-State Parameters and Braking Force Distribution

Abstract

In the study of braking force distribution of trucks, the accurate estimation of the state parameters of the vehicle is very critical. However, during the braking process, the state parameters of the vehicle present a highly nonlinear relationship that is difficult to estimate accurately and that seriously affects the accuracy of the braking force distribution strategy. To solve this problem, this paper proposes a machine-learning-based state-parameter estimation method to provide a solid data base for the braking force distribution strategy of the vehicle. Firstly, the actual collected complete vehicle information is processed for data; secondly, random forest is applied for the feature screening of data to reduce the data dimensionality; subsequently, the generalized regression neural network (GRNN) model is trained offline, and the vehicle state parameters are estimated online; the estimated parameters are used to implement the four-wheel braking force distribution strategy; finally, the effectiveness of the method is verified by joint simulation using MATLAB/Simulink and TruckSim.