Human-in-the-Loop Control Framework for Robot-Mediated Error Augmentation Training Based on Muscle Synergy Assessment

Robot-mediated error augmentation (EA) offers a promising paradigm for enhancing motor adaptation by amplifying movement errors rather than compensating for them. However, integrating EA into human-in-the-loop (HITL) control systems remains challenging. This difficulty arises from the lack of performance metrics that reflect neuromuscular adaptation and the limited ability to efficiently adapt control parameters to user-specific motor strategies. To address these challenges, we propose a HITL control framework that leverages muscle synergy analysis to guide the personalization of EA intensity. A synergy-based performance metric was designed to quantify motor adaptation by measuring structural similarity across muscle synergies. This metric served as the optimization objective for a Gaussian process-based Bayesian algorithm, which adaptively tuned the EA gain based on individual responses. An 8-day study involving ten healthy subjects performing non-dominant limb reaching tasks validated the proposed framework. The experimental group achieved significantly greater reductions in trajectory deviation than the control group, alongside robust within-group improvements ( ${p} \lt {0}.{01}$ ) and positive inter-group trends in muscle synergy similarities ( ${g}_{w}, {g}_{c}$ ). These results demonstrate reliable convergence of the optimization process, robust inter-subject adaptability, and the effectiveness of the synergy metric in capturing adaptation. Overall, the findings highlight the potential of the proposed framework as a foundational tool for adaptive motor training in human-robot interaction systems.