A personalized human–machine cooperative approach with transformer-based recognition for longitudinal and lateral control of intelligent vehicles

Human–machine interaction brings challenges for vehicle control design due to individual differences, a personalized cooperative approach with driving style recognition is proposed to achieve lateral and longitudinal control of intelligent vehicles in this paper. An improved Transformer-based method with an unsupervised pre-training and window-based multi-head self-attention is proposed to enhance the recognition accuracy and speed of driving styles, and thereby to capture the controller parameters under various driving styles. To achieve the lateral and longitudinal control of human-machine cooperative system, an integrated driver–vehicle model is established by considering driving styles and vehicle planar dynamics. Then, a Takagi–Sugeno fuzzy controller is developed to handle time-varying parameters and eliminate human-machine conflicts. Especially, stability conditions are exploited by Lyapunov arguments to achieve the control objective. Finally, simulation results show that the designed Transformer-based method has better classification accuracy and computational efficiency than other baselines on the same dataset. Based on recognition results, the designed controller can effectively improve the driving performance under various driving styles and time-varying parameters compared with other methods.