A Human–Vehicle Cooperative Path Following Control Method for Intelligent Vehicles

To address smoothness and safety issues in driving authority switching for human–machine co-driving intelligent vehicles, a hierarchical control strategy is proposed. The upper-layer controller utilizes a two-point preview model for human driving and a radial basis function neural network-based nonsingular fast terminal sliding mode controller with a heuristic gain adaptive tuning mechanism (RBFNN-HGATM-NFTSMC) for autonomous tracking. The lower-layer authority switching controller employs a driving risk-based finite state machine, integrating a risk-switching strategy and a cost function-based authority return mechanism. CarSim/Simulink co-simulations show that, compared to model predictive control, the upper-layer controller reduces maximum lateral errors by 77.7% and 66.2% under medium- and high-speed conditions, respectively. Compared with standard RBFNN-NFTSMC, it reduces lateral and heading errors by 56.2% and 11.9% at high speeds. With the decision logic held fixed, the risk-mapped transition lowers the takeover jerk peak by 35.75% and the return-phase jerk by 91.4% relative to an instantaneous transfer, at the cost of a larger transient lateral error during the handover. Under high-speed sudden distraction, the maximum lateral tracking error is 1.17 m. Under extreme parameter mismatch conditions, lateral and heading errors are reduced by 88.5% and 87.1% compared to standard NFTSMC, indicating robustness of the proposed controller under the tested parameter mismatch.