Human‐Machine Mutual Trust Based Shared Control Framework for Intelligent Vehicles

Due to technological bottlenecks in highly autonomous driving, human‐machine co‐driving will remain a transitional mode for the foreseeable future. As a promising paradigm, shared control allocates control authority adaptively by modeling interactions between humans and machines. However, existing methods often rely on rational game‐theoretic formulations that overlook psychological factors like trust, thereby limiting robustness to unexpected environmental variations. To address this limitation, this paper develops a shared control framework driven by bidirectional trust modeling between humans and machines. First, a human factors‐aware human‐to‐machine trust model is introduced, which derives trust estimates from intention discrepancies between the driver and the machine and refines them through driver‐skill assessments updated in a Bayesian framework. Second, a machine‐to‐human trust estimation model is introduced, which generates estimates from a comprehensive assessment of driving ability, integrating long‐term skills with short‐term states through a game‐theoretic mechanism. Finally, these trust estimates are mapped into control authority allocations and integrated into a model predictive control for generating shared control commands. Experiments with 25 drivers on the SILAB hardware‐in‐the‐loop platform validated the proposed framework. Results show high accuracy in trust estimation and demonstrate that the trust‐driven strategy achieves superior safety and human‐machine consistency compared with risk‐based approaches.