Self-tuning neural network generalized predictive control and energy consumption optimization for electro-hydraulic composite steering system of commercial vehicle

Heavy commercial vehicles equipped with traditional power steering system are inadequate to balance steering portability, steering stability, and fuel economy. Therefore, an electro-hydraulic composite steering (EHCS) system that integrates electric power steering (EPS) system and electro-hydraulic power steering (EHPS) system has been proposed. However, since the system has two actuators with different characteristics, how to effectively track assistance, reduce steering torque fluctuations, and further reduce energy usage has become a focus. This paper proposes a self-tuning neural network generalized predictive controller based on improved fruit fly optimization algorithm to address the above issues. Firstly, the dynamic model is derived through an analysis of the structure and dynamic properties of the EHCS system. To address the problem of energy usage, this paper develops a double-swarm division and cooperation fruit fly optimization algorithm (DDCFOA) to achieve optimal distribution of steering assist torque. The neural network implicit stair-like generalized predictive controller (NNISGPC) is designed to control the output assistance of a nonlinear time-delay EHPS system, and the issue of steering torque fluctuations caused by hysteresis is addressed through the adoption of electric power compensation control. Simultaneously, using DDCFOA to optimize the neural network and achieve controller parameters self-tuning. Finally, a joint simulation verification is conducted with AMEsim and Simulink. The results indicate that the suggested approach can effectively enhance the tracking performance of assist control, reduce steering torque fluctuations, and minimize energy usage.