To address the limitations of traditional mirror therapy in stroke rehabilitation, such as rigid movement mapping and insufficient personalization, this study proposes a robot-assisted mirror rehabilitation framework integrating multimodal biofeedback. By synchronously capturing kinematic features of the unaffected upper limb and surface electromyography (sEMG) signals from the affected limb, a dual-modal feature fusion network based on a cross-attention mechanism is developed. This network dynamically generates a time-varying mirror ratio coefficient $\lambda$, which is incorporated into the exoskeleton’s admittance control loop. Combining a trajectory generation algorithm based on dynamic movement primitives (DMPs) with a compliant control strategy incorporating dynamic constraints, the system achieves personalized rehabilitation trajectory planning and safe interaction. Experimental results demonstrate that, compared to traditional mirror therapy, the proposed system exhibits superior performance in bilateral trajectory covariance metrics, the mirror symmetry index, and muscle activation levels.

Multi-Agent Systems (MAS) excel at accomplishing complex objectives through the collaborative efforts of individual agents. Among the methodologies employed in MAS, Multi-Agent Reinforcement Learning (MARL) stands out as one of the most efficacious algorithms. However, when confronted with the complex objective of Formation Control with Collision Avoidance (FCCA): designing an effective reward function that facilitates swift convergence of the policy network to an optimal solution. In this paper, we introduce a novel framework that aims to overcome this challenge. By giving large language models (LLMs) on the prioritization of tasks and the observable information available to each agent, our framework generates reward functions that can be dynamically adjusted online based on evaluation outcomes by employing more advanced evaluation metrics rather than the rewards themselves. This mechanism enables the MAS to simultaneously achieve formation control and obstacle avoidance in dynamic environments with enhanced efficiency, requiring fewer iterations to reach superior performance levels. Our empirical studies, conducted in both simulation and real-world settings, validate the practicality and effectiveness of our proposed approach.