Human–Machine Cooperative Vehicle Control Based on Driving Intention and Risk Avoidance

This study explores human–machine collaborative planning and tracking control methods for autonomous vehicles. The proposed approach is based on an improved Fastformer GAN (FGAN) algorithm and a risk assessment mechanism for driver behavior. Initially, a novel trajectory prediction method that integrates the Fourier Attention Fastformer (FAF) and GAN models is proposed. This method is used for predicting driver behavior and adjusting trajectories based on driver intent, correcting any improper actions by the driver. Subsequently, a risk assessment system, which couples Artificial Potential Field (APF) and Dynamic Potential Field (DPF) models, is introduced to evaluate the risk levels of driver behavior. Adaptive activation of human–machine collaboration is based on the assessed driving risks. Simulation results indicate that the proposed FGAN algorithm significantly improves trajectory prediction accuracy on public datasets. Furthermore, the proposed human–machine collaboration method ensures both vehicle safety and stability while greatly reducing human–machine conflicts in real-time applications, demonstrating its feasibility and effectiveness.