Brain-Controlled Wheeled Mobile Robots: A Shared Control Framework Integrating Event-Triggered Mechanism and Deep Reinforcement Learning.

This study addresses the problem of quantifying user control authority in brain-computer shared control by integrating Event-Triggered Control (ETC) with Deep Reinforcement Learning (DRL). Firstly, an ETC-based brain-computer shared-control framework is developed for a wheeled mobile robot (WMR). In this framework, the Steady-State Visual Evoked Potential brain-computer interface (SSVEP-BCI) directly controls the WMR during non-triggered intervals, while control is transferred to a model predictive controller (MPC) once an event is triggered. Secondly, to overcome the limited adaptability of the Fixed Threshold (FT) triggering mechanism, a DRL-based adaptive triggering strategy is introduced to replace manually designed threshold rules. A grouped training strategy is further adopted to account for inter-subject differences in SSVEP-BCI decoding reliability during DRL training. Finally, experimental results demonstrate that integrating ETC into the SSVEP-BCI shared-control system improves the path-tracking performance of brain-controlled WMRs while enabling explicit quantification of user control authority. Specifically, compared with the FT-based strategy, the proposed DRL-based method achieves comparable lateral tracking performance, reduces heading error by 32.34%, and lowers intrusion rate by 57.85%. In addition, compared with the Time-Triggered Shared Control baseline, the cumulative execution time is reduced by 82.38%. These results indicate that the proposed framework achieves a favorable trade-off among tracking performance, computational cost, and preservation of user control authority.