当前位置: 首页>博士论文>资源详情
基于重复学习的下肢外骨骼控制研究
中文摘要

下肢外骨骼系统既能帮助脑中风、老龄化、肢体运动障碍以及残疾等人群提供康复训练或行走助力,亦能增加人体载荷能力,对改善人类生活条件与提高生活品质有重要作用。作为下肢外骨骼技术的核心,控制系统的优劣直接决定系统工作性能。良好的控制性能不仅能提升穿戴者的舒适度,而且对于保障训练质量或提升助力效率至关重要。下肢外骨骼机械系统为强耦合复杂非线性系统,其机构与伺服驱动系统均存在模型不准确、参数变化、外部扰动以及与人体相互作用等影响,控制难度大。 下肢外骨骼的工作条件使得系统具有重复或周期特征,具体表现为重复动态或重复扰动。对系统重复信息充分利用能在进一步提高控制性能的同时,使系统对不同穿戴者更具学习适应能力。论文从下肢外骨骼对参考轨迹的跟踪控制入手,以重复学习控制为理论基础,结合Lyapunov方法,分别针对下肢外骨骼机械连杆系统、液压驱动系统以及完整系统进行了控制研究,实现了外骨骼系统对参考轨迹的高精度跟踪控制。主要研究内容如下: (1)针对下肢外骨骼机械连杆系统,采用Lagrange建模方法,建立了下肢外骨骼系统模型。基于此模型,研究了下肢外骨骼重复学习控制。通过引入一种联合误差因子,实现了下肢外骨骼Lagrange系统快速跟踪控制。结合Lyapunov方法,在保证系统稳定性的同时,通过调整参数进一步提高参对考轨迹跟踪误差的收敛速度。针对外骨骼机械连杆系统的未知非线性与自身重复扰动特性,结合神经网络与重复学习控制理论,研究了下肢外骨骼神经网络重复学习控制。 (2)针对下肢外骨骼液压驱动系统,研究了其非线性重复学习控制。在考虑施加到液压缸活塞杆上等效力为周期性的同时,将不同摆幅条件下活塞杆所推动等效质量分成未知常量与未知周期变量两种情况进行处理。针对活塞杆所推动等效质量为未知常数的情况,结合投影映射自适应参数估计方法与重复学习方法,研究了液压驱动系统重复学习控制;针对活塞杆所推动等效质量未知且周期性变化情况,使用离散化技术,研究了液压驱动系统重复学习采样控制。 (3)实现了一套下肢外骨骼系统样机,分析计算外骨骼液压驱动系统参数,配置适用于下肢外骨骼的传感器系统;研制了用于捕获传感器数据的人体步态数据采集系统、用于实现控制算法的下肢外骨骼电子控制系统。 (4)以外骨骼完整系统为对象,研究了由液压驱动外骨骼系统完整动力学模型,并基于此模型研究了外骨骼重复学习控制。引入由周期参考轨迹产生的虚拟重复动态,将外骨骼完整系统实际动态与虚拟重复动态之差等效成控制系统上界,设计了非耦合重复学习控制器。特别地,基于前文所描述的外骨骼系统样机,将所设计的控制器实现到了数字电子系统并进行了实验,达到了良好的效果。 关键词:下肢外骨骼;重复学习控制;拉格朗日系统;液压驱动系统;跟踪控制

英文摘要

Lower-limb exoskeletons help stroke, aging, limb movement disorder and disabled persons in the aspects of rehabilitation training, walking assistance, or functionality enhancement of human body. Such kind of wearable robots plays a more and more important role in improving quality of life as well as capability of motion. As exoskeletons work in association with human motion, effectiveness of control determines directly performance of the human-machine system. A control scheme is adoptable if it not only makes wear confortable, but also ensures training quality and assistance efficiency. Dynamics of an exoskeleton is time-varying, strongly nonlinear and usually accompanied with coupling effects. Furthermore, the mechanical structure and its actuation system are both subject to model inaccuracy, parameter variation as well as external disturbances. During the walking process, the interaction with human body makes the control even more difficult. For rehabilitation and regular motion of the wearer, the lower-limb exoskeleton system possesses periodic dynamics accompanied with repetitive disturbances. It therefore makes sense to tacle the problems of uncertainty and disturbances in a framework of repetitive learning. This dissertation focuses on the trajectory tracking control problem of lower-limb exoskeletons, where the repetitive learning control theory and the Lyapunov method are two major tools utilized for analysis and controller design. The mechanical structure, the hydraulic actuator and the complete system of the exoskeleton are all considered, respectively. The main contributions presented in this dissertation are summarized as follows: (1)Based on the Lagrange approach, the model of lower-limb exoskeleton mechanical system is derived and its relevant repetitive learning controller is developed. By introducing a combined-error factor, fast tracking control of the exoskeleton system is studied. Combined with the Lyapunov method, the convergence speed of tracking error is addressed. Furthermore, a neural-network repetitive-learning controller is proposed to deal with both the non-periodic and periodic unknowns. (2)A novel repetitive learning control scheme of the hydraulic drive system of exoskeleton is proposed. The equivalent force acts on the piston rod of the hydraulic cylinder is considered as periodic. The equivalent mass applied on the piston rod is divided into two parts, namely an uncertain constant and an unknown periodic term. In order to resolve the unknown constant problem acted on the piston rod, an adaptive repetitive learning controller is proposed by utilizing adaptive parameter estimation. Based on Euler approximation, a sampled-data repetitive-leaming controller is proposed to address the unknown periodic term. (3)A prototype of lower-limb exoskeleton is developed. Calculation of the parameters of the hydraulic drive, layout and configuration of the sensors, a human gait data acquisition system as well as an embedded electronic control system are developed and tested. (4)The complete model of the lower-limb exoskeleton driven by hydraulic actuators is presented. By introducing a virtual repetitive dynamic term caused by the periodic reference, errors between the virtual and the actual dynamics are proven to be upper bounded. A repetitive learning controller is developed to compensate such error. Further, the controller is implemented and tested in the embedded electronic system. The fesibility of the proposed controller is verified by experiment. Key Words: Lower-Limb Exoskeleton, Repetitive Learning Control, Lagrange System, Hydraulic Driving System, Tracking Control.

作者相关
主题相关
看过该书的人还在看哪些书