Adaptive Robust Force Control Of A Walking Lower Limb Hydraulic Exoskeleton

In many unknown unstructured environments, such as steep hills, steps, and narrow aisles,heavy load transportation often can not be realized by wheel-type devices such as automobiles.Due to good adaptation to various extreme terrains of the leg structure, walking becomes an ef-fective means of transporting heavy loads in these situations.Although a lot of foot-type robots have been developed, in the short term, fully autonomous foot-type robots adapting to different ter-rains are still difficult to be achieved due to various technology limitations. As a typical wearable human-machine integrated device, lower limb exoskeleton introduces human into the robot con-trol which greatly simplifies the controller design of the robot. Compared to the fully autonomous foot-type robot, the lower limb exoskeleton avoid a lot of hard to solve technical limitations, and therefore has greater application potential. When the exoskeleton can track the human motion ac-curately, the heavy load is transferred to the ground through the exoskeleton and the human almost can not feel the existence of the load, and thus can complete various actions flexibly. Therefore,the high performance human machine synchronization control algorithm design becomes the key technology in the development of lower limb exoskeleton for human performance augmentation.However, the inherent nonlinearities, model uncertainties and multi-degree of freedom coupling as well as multi-gait characteristics bring great difficulties to the control algorithm design.This paper takes the hydraulic lower limb exoskeleton as the research object and the high precision human machine synchronization control as the research objective. With the help of theoretical analysis, simulation and experimental research, the high precision human machine synchronization control strategies of the lower limb exoskeleton are investigated systematically and thoroughly. First, as for the nonlinearities and model uncertainties in exoskeleton system, a 1-DOF adaptive robust cascade force control method is proposed to overcome the poor robust-ness problem. The effective inference of human motion intent as well as accurate tracking of the exoskeleton to the human movement have been achieved. In order to overcome the bandwidth limitation of cascade control, a 1-DOF adaptive robust backstepping force control based on the whole dynamic model is also proposed to achieve faster response for the exoskeleton. Secondly,as for the multi-joint coupling in the exoskeleton system, an observer based MIMO adaptive robust cascade control method is proposed. An adaptive robust observer is designed to estimate the joint acceleration effectively. Based on the estimated joint velocity and acceleration, a multi-variable low level adaptive robust motion tracking controller is designed. Combined with high level hu-man motion intent inference controller, the accurate tracking of a single leg exoskeleton to the human movement is ultimately achieved. Finally, as for the multi-gait and closed-chain dynamics in the lower limb exoskeleton, a general multi-gait dynamic model is established for the floating lower limb exoskeleton in which different holonomic constraints are added for different contact conditions of the exoskeleton foot. Then for single leg support and double leg support, the corre-sponding adaptive robust force controller is designed to minimize the human machine interaction force at the end effectors. Ultimately, an accurate walking of the exoskeleton with 20Kg load to human motion is achieved.The dissertation consists of the following six chapters:In Chapter 1, the research background of lower limb exoskeleton for human performance aug-mentation is detailed and the three control issues associated with high precision human machine synchronization control are pointed out. The domestic and foreign literature survey of exoskele-ton system, including the exoskeleton structures, human machine synchronization control methods and advanced motion control of robot manipulator are given. A brief summary of the dissertation's contributions and significance is subsequently given.In Chapter 2, the hardware equipment and their specifications used in the experiments are introduced. The kinematic model of lower limb exoskeleton system is established. The whole nonlinear dynamics of lower limb hydraulic exoskeleton is presented with focus on modeling of rigid body dynamics of exoskeleton structure, hydraulic actuators and human machine interface.Finally, the main parameters in the dynamic model are obtained through system identification and parameter calculation.In Chapter 3, as for single-joint exoskeleton system, a 1-DOF adaptive robust cascade force control method is proposed. In the high-level, the integral of human-machine interaction force is minimized to generate the desired position (which can also be seen as the human motion intent).And in the low-level, the accurate motion tracking of the generated human motion intent is devel-oped. The nonlinear high-order dynamics with unknown parameters and modeling uncertainties are built, and adaptive robust control (ARC) algorithms are designed in both control levels to deal with the complicated nonlinear dynamics and the effect of parametric and modeling uncertainties.Comparative simulation and experimental results indicate that the proposed approach can achieve smaller human-machine interaction force and good robust performance to various uncertainties. In order to overcome the bandwidth limitation of cascade control, a 1-DOF adaptive robust backstep-ping force control based on the whole dynamic model is also proposed. Comparative simulation results indicate that the proposed approach can achieve faster response for the exoskeleton.In Chapter 4, as for single leg exoskeleton, an observer based MIMO adaptive robust cas-cade control method is proposed to overcome the multi-joint coupling effect. A multi-variable high level controller is designed to infer the human motion intent. An adaptive robust observer is designed to estimate the joint acceleration effectively. Based on the estimated joint velocity and acceleration, a multi-variable low level adaptive robust motion tracking controller is designed.The effect of controller and observer gains on the control performance are analyzed theoretically.Comparative simulation results indicate that the proposed approach can achieve higher prediction than independent PID joint control.In Chapter 5, as for lower limb exoskeleton, a multi-gait adaptive robust cascade control method is proposed to overcome the multi-gait control problem. Based on the dynamic models established in the second chapter, the corresponding adaptive robust force controller is designed to minimize the human machine interaction force at the end effectors both in single leg support and double leg support condition. Ultimately, an accurate walking of exoskeleton with 20Kg load to human motion ia achieved.In Chapter 6, the research work of this dissertation is summarized. Major innovations are highlighted, and some future research directions are discussed.