容错型康复外骨骼生命机械体共融感知与决策机理研究

Exoskeleton is used to aid stroke patient lower limb early stage walking rehabilitation training. However, more attention should be paid on research of human-machine interaction, especially human-machine coupling joints interaction force measurement, gait trajectory planning with characteristics of both fault-tolerant and balance, and adaptation effect of patients to target gait trajectory during different stages. Through analysis of human-machine interaction force and walking balance model, combining with robust fault-tolerant adaptive rehabilitation training gait, friendly human-machine interaction can be realized. The project is carried out as follows: First, by analysis of human joint kinetic features, exoskeleton self-adaptive joint design parameters are obtained, with which human-machine coupling joint interaction force can be precisely measured; Then, by measuring human multimodal information using sensor system, the mapping model between human multimodal information and walking balance can be built, with the help of zero moment point theory and deep learning algorithm, and the target gait trajectory during each sub-stage can be formulated which also fits balance constraints; Last, based on physical therapist’s experience, stroke patients phased rehabilitation effect evaluation system is built, with which each sub-stage rehabilitation guidance process is updated combining with exoskeleton self-learning method, leading to more efficient and reasonable rehabilitation training assistant. Successful implementation of the project will provide researchers with relatively systematic exoskeleton human-machine communion perception and decision theory and method, and push forward clinical application of rehabilitation exoskeleton robots.

外骨骼可用于辅助脑卒中患者下肢早期步行康复训练，然而在人机交互方面，尤其是人机耦合关节交互力测量、具有容错平衡能力的步态轨迹规划以及患者对各阶段目标步态轨迹适应效果等方面的研究有待进一步深入。通过人机交互力分析和步行平衡模型建立，制定鲁棒容错自适应的康复训练步态是确保人机友好交互的有效手段。项目展开如下：首先，分析人体关节运动特征，设计外骨骼自适应关节，实现对人机交互力的精确测量；其次，利用传感系统检测人体多模态信息，结合零力矩点平衡理论和深度学习算法，建立人体多模态信息与步行平衡的映射模型，制定分级阶段符合平衡条件的目标步态轨迹；最后，基于康复专家经验建立脑卒中患者阶段性康复效果评价体系，通过外骨骼自学习修正各阶段康复指导进程，实现更高效合理的康复训练。本项目的成功实施将得到一套较为系统的外骨骼生命机械体共融感知与决策理论和方法，为推动康复训练外骨骼临床应用提供理论意义和实际价值。

下肢外骨骼机器人可有效辅助脑卒中、脊髓损伤患者实现早期的物理康复训练，也可以帮助老年人辅助行走延缓肌力衰退。然而，在人体运动意图检测、步态规划与外骨骼力、位置控制方面存在着相关理论研究欠缺，人机物理交互和感知交互有待提升。本项目通过研究外骨骼结构优化设计、人体步行意图和步态相位识别算法以及步态规划和助力策略相关内容，有效解决了下肢外骨骼机器人在辅助人体步行过程中的平衡安全性问题，人机深度共融的助力协调交互问题，并开发出分别用于下肢辅助步行康复训练和步行助力的下肢外骨骼机器人实验系统。前者利用基于组合足底压力中心算法，通过迈步落脚点计算模型合理规划了平衡步态，实现了综合人机状态的平衡步行，实验结果表面该算法可良好适应人体在重新前后变化时迈步落脚点的自适应变化，为实现下肢医用康复外骨骼推广安全使用提供了技术保障；后者借助穿戴式传感器提取人体步行过程中髋关节角度，借助本研究提出的混合振荡器模型对步态相位进行了连续精准识别，并作为助力矩生成的输入信号，最终实现对人体自由步行的良好适应和合理助力，为解决助行外骨骼在日常生活场景的使用提供了理论支撑。本项目上述研究成果验证了所提出的人机共融理论研究的可行性和有效性，形成了项目预期计划的外骨骼生命机械体共融感知和决策的理论和具体实现方法，推动了康复训练外骨骼在实际的应用。