适应复杂环境的主动型假肢建模与平稳控制机理研究

In order to overcome the drawbacks of unsteady gait during walking, a strategy of identifying and steady control for powered prosthesis users in a complex environment is proposed in this project. The main contents are as follows: A novel thought about the judgment of human motion intention based on the fast orthogonal search method is proposed by use of the motion signals. Then a motion intent database is constructed. From the point of view that the amputees and the prosthesis constitute a man-machine system, based on the improved neural networks (networks' structure, networks' learning algorithm), the rules of choosing a suitable control gain to improve the steady of prosthesis users during walking in different speeds and on different terrains or paths are discussed preliminarily. For this man-machine system with the input and output constraints, a nonlinear predictive control algorithm based on piecewise function is proposed. The corresponding control constraints are determined according to the elliptical stability domain where the current states are lying in. By offline designing the piecewise predictive control law and online implementing the appropriate predictive control law, the online computational efficiency is improved greatly. .These works provide the necessary theoretical basis and experimental data for development of powered prosthesis. With improvement of the related technologies, some products on the powered prosthesis are realized. This is positive academic significant and practical important to improve the quality of life for amputees

本项目针对主动型假肢在使用过程中存在的步态不平稳问题，研究复杂环境下人体运动状态发生改变时，使用者运动意图建模及假肢平稳控制。研究内容主要包括：根据人体运动信息，采用快速正交搜索法研究人体运动意图建模新思路，并建立相应数据库；提出改进型神经网络（结构、学习算法）等，实现在不同步速、不同路况下，假肢使用者步态平稳控制增益选取规则；根据运动意图模型的输入输出受约束问题，实行分段划分，根据状态所处的不同稳定椭圆域确定其对应的控制约束，设计离线型预测控制器，使得预测控制的主要计算都离线完成，在线不涉及优化问题，实现人机交互的主动型假肢平稳控制。.通过对主动型假肢关键技术的研究，为发展面向应用的智能假肢系统，提供必要的理论依据、实验数据。随着相关技术不断发展完善，将主动型假肢技术转化为产品，这将对于提高截肢者的生活质量，具有积极的学术意义和重要的实际意义。

本项目研究了复杂环境下主动型假肢的平稳行走问题。针对主动型假肢，研究了基于智能计算方法的人体运动环境识别方法，根据采集的人体信息，基于极限学习机、概率神经网络和BP_Adaboost分类算法等实现了人体外界环境识别，通过仿真实验证明所提方法的有效性。进而针对假肢的实时性、路况识别准确率要求高等特点，提出了一种基于激光测距传感器的人体运动环境识别系统，借助激光测距传感器和惯性测量单元完成了地形测量，通过概率神经网络、随机森林算法和BP_Adaboost分类算法实现了人体运动意图预测，所提方法证明了具有较高的精确度和实时性。针对建立的主动型假肢模型设计离线型预测控制器，预测控制的主要计算都离线完成，在线不涉及优化问题。提出基于自抗扰控制技术的主动型假肢控制方法，将各种扰动组合统一观测并统一补偿，有效抑制作用于主动型假肢的各种扰动，从而获得高性能的主动型假肢控制系统。对所提算法设计了人体路况仿真实验，并在主动型假肢中进行了实验验证，控制效果好，较好地实现人机交互的主动型假肢平稳控制。本项目中所提方法现在还处于实验阶段，对把主动型假肢技术转化为产品具有积极的指导意义和重要的实际价值。