Human-robot Coexisting Strategies Of Lower Extremity Augmentation Exoskeleton Control System

The lower extremity augmentation device is considered as a wearable robotic device.The development of this kind of robots has made the weight-bearing locomotion,lifting heavy gear as well as frontier inspection and patrolling possible without reducing the human movement ability.Thus,the working efficiency and battle capability can be improved.The lower extremity exoskeleton can organically integrate human body parts,usually consisted of two anthropomorphic legs,upper-body backpack,and corresponding facilities.The exoskeleton is mainly used to reinforce the pilot's endurance,improve his weight-bearing capability and protect his movement limbs.This dissertation aims to contribute the improvement of coexisting capability between the lower extremity exoskeleton and the user by studying the human-robot coexisting strategies of the control system of the lower extremity augmentation exoskeleton.Firstly,we carefully analyse the special issue of the joint motion of the human lower-limb and the normal locomotion derived from the human biomechanics and human anatomy.The dynamic model of the swing phase and stance phase are introduced in this dissertation,along with the dynamic system based on hybrid automata of the lower extremity exoskeleton.The control systems of the flexible-joint exoskeleton system,lightweight exoskeleton system as well as heavy-weight exoskeleton system are designed.Secondly,the interaction model based on the sparse Gaussian Process and central pattern generator for rhythmic movement primitives are built.According to the Policy Improvement and Path Integrals theory and CMA-ES algorithm,a novel human-robot interaction incremental movement primitives learning algorithm is proposed.In addition,the weight parameter adaptation for the controller is given.Thus,a novel human-robot coexisting strategies for high-level movement primitives on-line reinforcement learning based on the Position Loop.Moreover,derived from the Hidden Markov Model and Gaussian Process,the identification system model,which is used to identify the state space variables,is introduced with Gaussian Process filtering and smoothing.Combining the on-line reinforcement learning algorithm(1 + 1)-CMA-ES,the sensitivity amplification algorithm is modified in order to adapt the sensitivity amplification parameter.Four data fusion methods aretested for completing distributed scheme.Therefore,based on the Force Loop,a novel probabilistic sensitivity human-robot coexisting strategies is proposed in this dissertation.Finally,the dynamic model of the series elastic actuator is built.Derived from the intrinsic sensing of the series elastic actuator,the interaction between the lower extremity augmentation exoskeleton can be seen as the disturbance and compensate to the controller based on the Single Joint Level.By designing the Q filter,the system robustness is thus enhanced.Combining with the model predictive control,a novel model predictive human-robot coexisting strategies based on the disturbance for the series elastic actuator of the lower extremity augmentation exoskeleton is proposed.Furthermore,owing to the properties of the internal model control,the local actual internal model can hence be built with on-line Distributed Gaussian Process Learning.Through on-line data flow,the modelling precision of the forward model can be guaranteed.Therefore,a novel evolving internal model human human-robot coexisting strategies is introduced for the series elastic actuator of the lower extremity augmentation exoskeleton.