面向偏瘫患者步行康复训练的柔索机器人动态规划与控制方法研究

Hemiplegic patients after stroke suffer from functional disabilities of the affected side and show abnormal gait during walking. This project aims to develop a rope-actuated robotic system to assist the locomotor rehabilitation training of hemiplegic patients. The robot provides supportive force in the three direction of up, forward, left and right, and thus can assist the patient with body weight support, push forward, gait transition, balance, etc. However, the patient must initiate the walking by himself/herself. In this way, the robot provides personalized assist-as-needed support to the patients, while avoids the slacking effect. First of all, we analyze the abnormal gait of hemiplegic patients, and build models to predict the gait with measured ground reaction force, kinematic and physiological data. Then, we study the relationship between the assistive force, ground reaction force, kinematic data with the resulted gait, and thus can dynamically plan the assistive force for the desired gait. Afterwards, we improve the control performance of the cable-driven series elastic actuators, analyze dynamic properties of the overall rope-robot system, and design the control scheme to achieve precise control of the highly dynamic assistive force. Then, we will build up the rope-robot system, and validate the developed methods by experiments with healthy people and patients. Thus, a feasible solution will be developed for robot-assisted locomotor rehabilitation of hemiplegic patients after stroke.

本项目面向脑卒中后一侧肢体有显著功能障碍的偏瘫患者，采用柔索驱动，施加向上、向前和左右三个方向的拉力，为其提供负重支撑、前行动力、步态切换和身体平衡的辅助，而行走则必须由患者主动进行。这既基于患者的运动障碍和康复需求提供了“个性化按需辅助”，又从根本上避免了“惰性效应”。具体而言，融合患者运动学、肌电、地面反作用力等多模态信息，分析其运动障碍，建立模型预测其异常步态。采用机理分析和灰箱建模方法，探究辅助力、地面反作用力和步态的关系，建立辅助力的动态规划方法。采用柔索牵引串联弹性驱动，基于多目标鲁棒控制，优化柔索驱动单元的性能。分析柔索机器人的非线性耦合特性，采用Lyapunov方法设计系统整体控制方案，实现高动态辅助力的精准控制。建立柔索机器人康复实验平台，通过仿真、健康人和病人实验，验证各种方法的有效性和实际步行辅助效果。由此，将为脑卒中偏瘫患者的步行康复训练提供一个有效的解决方案。

我国老龄人口已超过2.6亿，失能半失能人口达4000万。脑卒中是我国成年人致残第一大原因，步态异常是卒中患者的主要功能障碍，也是影响日常生活能力和生活质量的主要因素。本项目基于脑卒中致残机理和脑可塑性康复原理，研究机器人辅助步行康复训练技术与系统。提出基于血氧动力学的脑功能状态转换特征，设计脑功能区域拆解算法，突破EEG-fNIRS时频特征融合与协同分类技术，建立了脑功能和障碍的量化分析方法。设计肌电信号特征引导的卷积神经网络实现下肢运动意图精准估计，突破面向异常步态检测的步态视频和足底压力分析技术，建立了肢体行为功能和障碍的量化分析方法。设计柔性驱动的非线性控制方法，构建环形气动执行器的主动建模与控制技术，提出安全、鲁棒的人机交互控制方案，建立了柔性驱动单元的非线性运动控制和鲁棒人机交互控制架构。建立了提供视觉辅助的被动型步行训练辅助机器人和主动康复的步态与平衡训练机器人系统，并开展性能测试。开展临床试验，证明了机器人辅助下肢康复训练的有效性和安全性。项目成果为进一步建立临床实用化步行康复训练系统提供了坚实的理论和技术基础。