Adaptive Iterative Learning Kalman Consensus Filtering for High-Speed Train Identification and Estimation

Abstract

In this study, a data-driven adaptive iterative learning Kalman consensus filtering (DD-AILKCF) method is designed for high-speed trains to address the parameter identification and speed consistent optimal estimation problem. The nonlinear train dynamics model is transformed into a linear-like state-space model by using the Full Form Dynamic Linearization (FFDL) technique. Meanwhile, four types of sensors are used to obtain different kinds of datasets to implement the multi-sensor system. The method proposed in this paper consists of two steps. First, an adaptive iterative learning Kalman filtering (AILKF) algorithm is proposed to estimate the fast-time varying train parameter in the iteration domain. Then, based on the identified parameter, a distributed multi-source heterogeneous network consensus filtering (MHN-CF) algorithm is proposed for the speed estimation of high-speed trains. The convergence of the proposed algorithm is derived based on the Lyapunov Function. The proposed method is compared with existing methods by numerical simulations, and the results indicate that the proposed method achieves good effectiveness in improving the accuracy of high-speed train speed estimation.