人机自然交互是实现共融机器人“以人为本”的关键要素。然而，在面对动态环境和复杂运动任务时，仅依靠机器人拟人化的结构设计及智能控制方法，无法有效实现人机协调运动和柔顺动作。而人机之间的运动失调、动作失误，还可能导致机器人人机系统中的本体出现能量衰竭甚至机能受损等问题。本项目拟基于人体运动生理学实验研究，揭示复杂动态环境/任务下人体上肢运动、力的传递关系，提取肌肉协同与运动协同特征，建立融合肌肉激活、肌肉收缩和关节动力学特性的多关节肌骨骼动力学模型；基于肌骨骼模型和机器学习相结合的方法，实现动态环境下上肢功能性运动的多模态运动意图和运动状态同步估计，并通过相应的评估方案和评估指标对人机自然交互的有效性和可行性进行评估，为实现机器人人机交互系统在完成动态任务时的人机自然交互和对人本体的运动跟随与刚度匹配奠定理论基础。

Human-robot interaction (HRI) plays an important role in achieving “human-centered” movement in Tri-Co robotics. However, it’s not sufficient to obtain coordinated and compliant movement under complex environment and tasks only relying on anthropopathic mechanism design and advanced control theory. Moreover, the uncoordinated, noncompliant movement may lead to the energy failure and functional impairment. This proposal will start from human kinematic and physiologic experimental research to explore and analyze the variation tendency of kinematics, stiffness and muscle activities with variations of tasks and environments. The muscle synergy and motion synergy will be extracted and the dynamic model of musculoskeletal system will be built considering muscle activation, muscle contraction and dynamic characteristic of joint. The human movement intent and movement states will be simultaneously estimated from the combination of the musculoskeletal dynamic model and machine learning algorithm during functional upper limb movement under dynamic environment. Some evaluation strategy and indexes will be developed to evaluate the effectiveness and feasibility of the proposed natural HRI methods. The application of HRI is hopefully extended to functional movement estimation under dynamic environment benefiting from the proposed research.