基于机理-数据混合模型的FES手功能康复系统迭代学习控制

功能性电刺激(FES)手部康复系统能够尽早介入康复治疗、缓解手功能障碍、恢复手部自主运动能力。但FES手功能康复系统存在模型不精确、控制精度低的瓶颈，无法满足手功能康复的需求。本项目针对FES手功能康复系统的瓶颈问题开展研究,具体研究内容为：融合机理和数据模型建立手功能康复系统串级混合模型，研究模型动态更新方法；建立混合模型的卷积矩阵表达，构建基于模型的非线性范数最优迭代学习控制(NOILC)，融合NOILC和模型预测控制建立变权重迭代学习算法，提出动态轨迹生成和基于策略的遗忘机制，解决康复运动迭代学习控制的初始位置不确定和非周期性主观运动干扰的问题；考虑输入梯度约束，建立柔性康复运动控制方法，提高康复运动控制精度和舒适度。最后，开发FES手功能康复平台，测试和完善所提出的控制策略。本项目研究将突破FES手功能辅助康复的瓶颈，为手功能康复提供理论支撑和解决方案。

Functional electrical stimulation (FES) based hand rehabilitation system could be involved in the rehabilitation as earlier as possible to relieve hand dysfunction and recover the voluntary movement ability. However, due to the shortcomings in FES-based hand rehabilitation system, such as inaccurate modeling and low control accuracy, the requirements of hand rehabilitation are not met. Aiming at the bottlenecks of FES hand rehabilitation system, a hybrid model of hand rehabilitation system is established by fusing mechanism and data model, and the model updating method is studied. The convolution matrix of the hybrid model is set up, based on which a model-based nonlinear norm optimal iterative learning control method (NOILC) is founded. Then, a variable weight iterative learning control algorithm is proposed by fusing NOILC and model predictive control. To deal with initial value uncertainty and aperiodic voluntary movements, a trajectory dynamic generating method and a forgetting mechanism are proposed. To improve the control precision and comfort level of rehabilitation movements, a flexible rehabilitation movement control method is presented based on the constraint of input gradient. In the end, a FES-based hand rehabilitation platform is built and the proposed methods are verified. The research will break through the bottlenecks in FES-based hand rehabilitation system, provide theoretical and technical support for FES-based hand rehabilitation.