融合模糊 ΔQ 学习的脑控机器人共享控制策略

打开文本图片集

中图分类号：TP242.6 文献标志码：A 文章编号： 1000-5013（2026）01-0093-11

Abstract：To address the issue that existing fuzzy logic-based shared control methods overly rely on expert experience，a shared control strategy for brain-controlled robotics integrating fuzzy Q -learning is proposed. The method combines fuzzy logic with reinforcement learning，enabling adaptive optimization of human-machine control weights.By designing areward-penalty function that feeds real-time environmental information back into the system，the human-machine weights are dynamically optimized based on human brain fatigue level and environmental complexity information. The resulting weights are used as coeffcients for synthesizing the direction vectors. Comparative experiments against traditional fuzzy logic-based methods demonstrate that the proposed approach enables real-time adaptive adjustment of human-machine control weights under complex environments，significantly improving trajectory smoothness and task completion eficiency. These results validate thefeasibilityand effectiveness of the proposed shared control strategy for brain-controlled robotic systems.

Keywords：steady-state motion visual evoked potential；brain-computer interface；shared control;fuzzy Q learning；mobile manipulator

脑机接口（BCI）是一种不依赖外周神经和肌肉组织，直接利用大脑信号实现与计算机或外部设备交互的技术[1]，为严重运动障碍患者提供了通过意念控制外部设备的新途径。（剩余14157字）