Objective To propose a motion performance-based adaptive assistance strategy to balance the differences in movement intensity caused by varying motor abilities in dual-person upper limb rehabilitation training.Methods A competitive interactive task suitable for dual-person rehabilitation training was designed.Two kinematic evaluation metrics were introduced,and fuzzy logic was applied to comprehensively assess the patients'motor performance.An adaptive control system framework was constructed to dynamically adjust the robot's assistance level through a difference-based adaptive step-size piecewise function.Four healthy persons were recruited ran-domly to test the system.Results The proposed strategy could adaptively adjust the level of assistance based on the participants'motor perfor-mance,balancing the differences in motor abilities between them and enabling both parties to engage at a similar competitive level.Conclusion The motion performance-based adaptive assistance strategy and adaptive control system framework may bal-ance the skill levels between participants with different motor abilities,avoiding the imbalance of training intensi-ty during dual-person interactive rehabilitation.