A Vehicle Driving State Estimation Algorithm based on Elman Neural Network and Unscented Kalman Filter

Vehicle driving states in a dynamic vehicle model are with strong uncertainty and nonlinearity. Simply applying traditional Unscented Kalman Filter (UKF) or Neural Network cannot meet the requirements of future vehicle driving state estimation, which may lead to a severe failure in the vehicle's electronic controlling system. This paper proposes an optimized algorithm based on Elman Neural Network (ENN) and UKF. Combining the superiority of ENN dealing with nonlinearity and the ability of UKF dealing with uncertainty, the proposed ENN-UKF significantly reduces the filtering error. Based on the nonlinear seven-degree-of-freedom vehicle model and Dugoff tire model, ENN-UKF is tested on standard Double Lane Change road conditions. The results demonstrate that our approach performs much better than traditional UKF with an improvement of at least 40%. This study indicates the feasibility of applying ENN-UKF in the vehicle driving state estimation.