TY - JOUR
T1 - An Online Bi-LSTM Modeling Method for Steering Feedback Torque of Steer-by-Wire System Considering Off-Road Conditions
AU - Zhu, Runzhuo
AU - Yang, Chao
AU - Zhang, Yuhang
AU - Gao, Yipeng
AU - Qie, Tianqi
AU - Wang, Weida
AU - Yan, Qingdong
N1 - Publisher Copyright:
© 2026 IEEE.
PY - 2026
Y1 - 2026
N2 - Steer-by-wire (SBW) systems replace mechanical linkages with electronic control, eliminating direct road feedback to drivers. This creates safety risks as the steering decisions of drivers lack road information. To avoid this, the steering feedback torque (SFT) is generated by a feedback motor to simulate road information. However, road feel cannot be directly measured. This makes accurate SFT modeling crucial, especially in off-road conditions where the randomness of the terrain greatly increases the difficulty of modeling. To address this issue, an online bidirectional long short-term memory (Bi-LSTM) SFT modeling method is proposed, considering off-road conditions. First, a nonlinear dynamic model of SFT with temporal characteristics is used to analyze its mechanism and screen out the influencing factors. An offline long short-term memory (LSTM) model for SFT is established. Then, in order to adapt to changing off-road conditions, a dual-module parallel scheme is proposed. The training module employs sliding-window data selection and periodic gradient-descent updates. The testing module uses updated parameters for real-time SFT prediction. On this basis, the online Bi-LSTM model for SFT is obtained, and the bidirectional mechanism is introduced to reduce the influence of noise on the model. Finally, the dataset is built using the SBW system test bench. The proposed online Bi-LSTM model is verified through a test bench and a real vehicle equipped with the SBW system. During accuracy verification under off-road conditions, the online Bi-LSTM model improves SFT accuracy by 33.33% and demonstrates enhanced real-time performance. The proposed method enables novel multicondition SFT modeling and cross-scenario transfer.
AB - Steer-by-wire (SBW) systems replace mechanical linkages with electronic control, eliminating direct road feedback to drivers. This creates safety risks as the steering decisions of drivers lack road information. To avoid this, the steering feedback torque (SFT) is generated by a feedback motor to simulate road information. However, road feel cannot be directly measured. This makes accurate SFT modeling crucial, especially in off-road conditions where the randomness of the terrain greatly increases the difficulty of modeling. To address this issue, an online bidirectional long short-term memory (Bi-LSTM) SFT modeling method is proposed, considering off-road conditions. First, a nonlinear dynamic model of SFT with temporal characteristics is used to analyze its mechanism and screen out the influencing factors. An offline long short-term memory (LSTM) model for SFT is established. Then, in order to adapt to changing off-road conditions, a dual-module parallel scheme is proposed. The training module employs sliding-window data selection and periodic gradient-descent updates. The testing module uses updated parameters for real-time SFT prediction. On this basis, the online Bi-LSTM model for SFT is obtained, and the bidirectional mechanism is introduced to reduce the influence of noise on the model. Finally, the dataset is built using the SBW system test bench. The proposed online Bi-LSTM model is verified through a test bench and a real vehicle equipped with the SBW system. During accuracy verification under off-road conditions, the online Bi-LSTM model improves SFT accuracy by 33.33% and demonstrates enhanced real-time performance. The proposed method enables novel multicondition SFT modeling and cross-scenario transfer.
KW - Long short-term memory (LSTM) method
KW - off-road conditions
KW - online learning method
KW - steering feedback torque (SFT)
KW - steering-by-wire system
UR - https://www.scopus.com/pages/publications/105031938300
U2 - 10.1109/TIM.2026.3670509
DO - 10.1109/TIM.2026.3670509
M3 - Article
AN - SCOPUS:105031938300
SN - 0018-9456
VL - 75
JO - IEEE Transactions on Instrumentation and Measurement
JF - IEEE Transactions on Instrumentation and Measurement
M1 - 7502510
ER -