Ikram, Mubasher (Shenzhen Institute of Advanced Technology, Chinese Academy of Sc), liu, li (Changsha University of Science and Technology), Shen, Qiguang (Guangxi University of Science and Technology), Zhang, Qingguang (Shenzhen Institute of Advance Technology, China), Liu, Jia (Shenzhen Institute of Advanced Technology Chinese Academy of Sci), Xu, Kun (Shenzhen Institute of Advanced Technology, Chinese Academy of Sc)

Data-Driven Approach for Accurate Wheel Slip Ratio Estimation in Electric Vehicles: Performance Comparison and Validation Analysis

Scheduled for presentation during the Poster Session "Safety and Reliability Techniques for Autonomous Vehicles" (ThBT15), Thursday, September 26, 2024, 14:30−16:30, Foyer

2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), September 24- 27, 2024, Edmonton, Canada

This information is tentative and subject to change. Compiled on October 8, 2024

Abstract

Data-driven techniques advance accurate vehicle state estimation for electric vehicles (EVs), especially for distributed drive electric vehicles (DDEVs) that require precise slip ratio estimation for safe motion control. This study proposes advanced data-driven estimation models of wheel slip ratio using machine learning (ML) techniques, i.e., long short-term memory (LSTM), gated recurrent unit (GRU), and extreme learning machine (ELM), to manage the dynamic complexities of EVs effectively. The proposed models process measurable input parameters and are trained on a comprehensive dataset comprising both synthetic and real-vehicle datasets. Comparative and sensitivity analyses evaluate each model's accuracy, robustness, and generalization capabilities. The results reveal highly accurate slip ratio estimation across diverse driving conditions. The LSTM and GRU models demonstrate exceptional performance even under low friction conditions, achieving variance accounted for (VAF) scores of 99% and 98% respectively. Meanwhile, the ELM model proves effective on high adhesion surfaces with an average VAF around 95%. Validation results on real-world data show the GRU model outperforms the others, achieving a VAF of 90%. This study advances vehicle state estimation by offering a robust, data-driven approach incorporating comparative and validation analyses.