非结构环境下外骨骼机器人主动意图理解与柔顺控制研究

Wearable exoskeleton robots have greatly improved the ability of indoor activity in patients with joint disease, but their actions in outdoor non-structural environments are still severely limited.Outdoor non-structural environment such as loose, gravel, convex-concave, obstacles terrain will make the wearer suffer uneven force at the end of foot, and prone to stumble or even instability, which also put forward a serious challenge to exoskeleton robot’s active intention understanding and compliance control. This project focuses on the interdisciplinary research of two scientific issues between the human move intention extraction, identification, understanding and power-assist under non-structural environment. The basic and experimental research of the key technologies, such as the motion feature extraction and clustering analysis based on EMG signal, the power-assist adjustment and allocation strategy in unstructured environment, the compliance control of joint force-position, and the system energy efficiency regulation. Besides, the understanding method of the active motion intention in unstructured environment has been developed, which contributes to establish the mapping relation of motion behavior pattern matching. Furthermore, the joint power distribution model under terrains such as soft, gravel, convex-concave and obstacles is established, along with the power-assist optimization strategy. Finally, the method of motion-force compliance control based on intention understanding and the optimization method of system energy efficiency in periodic motion are proposed. Adding up all the above works, a new research on theory, method and technology has been developed, which is benefit for the field applications of the wearable active-assisted exoskeleton robot.

穿戴式外骨骼助力机器人已大大改善关节疾病患者的室内活动能力，然而他们在室外非结构环境下的行动仍受到极大限制。室外非结构环境如松软、碎石、凸凹、障碍等地形会使穿戴者足端受力不均，易发生磕绊甚至失稳，这也对外骨骼机器人的主动意图理解和柔顺控制提出了严峻挑战。本项目围绕人体运动意图特征提取与辨识理解和非结构环境下助力柔顺控制两个科学问题进行多学科交叉研究，开展基于肌电信号的运动特征提取与聚类分析、非结构环境下助力调整与分配策略、关节力-位柔顺控制、系统助力能效调控等关键技术基础研究和实验研究。提出非结构环境下的主动运动意图理解方法，建立运动行为模式匹配映射关系；建立松软、碎石、凸凹、障碍等地形下的关节助力分配模型，提出优化助力策略；提出基于意图理解的运动-力柔顺控制方法，以及周期运动过程中的系统能效优化方法。为穿戴式外骨骼助力机器人走向野外实用提供新理论、新方法、新技术。

穿戴式助力外骨骼在非结构环境下的助力效果和穿戴舒适性是外骨骼机器人走向实用的前提，也是当前可穿戴式外骨骼研究的前沿。本项目针对可穿戴式外骨骼在非结构环境下的运动特征提取以及柔顺控制问题，采用理论分析、数字仿真及实验验证相结合的方法，对非结构环境下的生物力学特征提取与分析、负重情况下的人体动力学建模、非结构环境下穿戴者运动意图识别、控制体系结构等关键理论和技术开展深入研究，取得以下创新性成果：.1）完成非结构环境下人体生物力学的特征提取与数据集构建。将传统的人体运动工况扩展到非结构环境，对蹲起、上下楼梯、步行、跑步、跨沟、上下斜坡、搬移托举等工况进行数据采集，并对人体下肢（上肢）各关节及肌肉进行生物力学分析，构建了7大类32项典型工况共960组有效测试数据集，为外骨骼设计提供有效数据支撑。.2）将传统无负重人体动力学建模扩展到有负重情形，基于拉格朗日方法建立人体蹲起和起跳的动力学模型，完整描述了负重情形下人体下肢各关节的力矩及功率，为助力外骨骼的详细设计提供依据。.3）针对sEMG信号原理及特征，提出了采用50Hz工频陷波和10-500Hz带通滤波的预处理方法，进一步提出了均方根+滑窗+圆滑处理的sEMG信号特征提取方法。基于BP神经网络算法建立了人体关节角度预测模型，提出适用于非结构环境下的穿戴者运动意图识别方法。.4）针对穿戴式外骨骼与人体肢体运动的耦合问题，建立了基于OpenSim的耦合运动学/动力学仿真，实现了穿戴式外骨骼关键几何参数与人体肢体运动的合理匹配。.5）研制了适用于典型工况的下肢助力外骨骼座椅和上肢肩关节助力外骨骼样机，研发了基于薄膜压力传感器的足底多点压力鞋垫，完成了诸如上下阶梯、上下斜坡、修理、装配、钻孔、测量、上/下肢协同等典型工况下的测试试验。相关成果获得陆军装备部“超能勇士-2019”单兵外骨骼系统挑战赛西安赛区“跨越障碍”项目第一名。