Research On Human-robot Cooperative Control Of Loosely-coupled Lower Limb Load-carrying Exoskeleton Robot

Augmentation exoskeleton robot has the advantage of human-robot complementation.The robot has advantages of carrying heavy load,high torque output and strong support ability.The human has superiorities of decision-making and flexible movement ability.Exoskeleton system combines the advantages of both of them.The man-machine system not only can operate in complex terrain freely,but also has higher obstacle crossing ability and load-carrying capacity.The system is one of the hot research directions of the robot in recent years.However,the difficulty of the exoskeleton system lies in the human-robot combination as well.The design of the control system needs to consider human motion intention recognition,human safety and comfort,and assist effect evaluation,etc.The control of traditional augmentation exoskeletons mainly focuses on the power goal,neglecting the dominance and actual feeling of the wearer,which is likely to bring other adverse effects to the wearer or the system,and fails to reflect the advantages of human-robot integration.This thesis selects loosely-coupled lower limb load-carrying exoskeleton as the research object,and proposes a human-robot cooperative control method.It makes the exoskeleton take into account both the comfort and stability of the system under the premise of bearing heavy load,and finally realizes the purpose of reducing the human-machine interaction and enhancing the stability of the system during walking.The main contents and innovations of this thesis are as follows:The human-robot interaction characteristics of the exoskeleton system are studied.Based on the factors of mechanical structure design,human-robot motion interference and device wearability,the concept of loosely-coupled exoskeleton is proposed and a typical loosely coupled lower limb load-carrying exoskeleton robot is designed.According to the different of single leg support phase and double leg support phase,the seven-link and four-link dynamics models in the sagittal plane of the exoskeleton were established respectively based on Lagrange method.Based on the interaction between human and environment,a physical human-robot interaction model of spring-damping is established.In order to solve the problem of the inherent uncertainty of human movement when the wearer is tracked by the lower extremity load-carrying exoskeleton,a method of predicting human movement trajectory based on probability model is proposed.Based on the human body horizontal ground uniform walking gait of rhythmic and uncertainty,the movement of the human barycentric position is regarded as a gaussian process by using statistical methods.Multidimensional gaussian function is adopted to establish the human body centroid trajectory model.Using the basic principle of iterative learning,the model is extended to variable speed case,so that the gait trajectory prediction can be achieve.In addition,the human-robot system is a two-body system.Aiming at the problem of human-robot motion bifurcation,the multi-dimensional Gaussian function based on human-machine interaction force is adopted to model the motion bifurcation,so as to facilitate the design of the human-robot cooperative controller.Aiming at the friendly assist problem of the loosely-coupled lower limb load-carrying exoskeleton,a human-robot cooperative control method based on comfort and stability constraints is proposed,which is variable factor risk sensitivity model predictive control considering the motion uncertainty of human.The Mahalanobis distance is extracted based on the human motion behavior model,and the weight of the cost function in the optimal control is adjusted by using the mahalanobis distance,thus the uncertainty of human motion is introduced into the controller to reduce the human-robot interaction force.Based on the risk sensitive optimal control,the nonlinear mapping relationship between zero moment point and risk factor is designed according to the system state,thus the stability of the system is introduced into the controller.Finally,the effect of reducing human-robot interaction force when the system stability margin is high and the position of zero moment point is adjusted when the system stability margin is low is achieved.Aiming at the problems of system modeling error,external interference and load fluctuation in the design of the low-level controller,an adaptive robust controller based on the neural network of fuzzy cerebellar model articulation controller is proposed.The objective of the controller is to provide an accurate desired driving force for the upper human-robot cooperative control.The desired driving force is transformed into the desired trajectory of the joints,and a computed torque controller is designed to track the desired trajectory.The fuzzy cerebellar model is used to approximate the equivalent modeling error part to ensure the convergence of trajectory tracking.Finally,the proposed human-robot cooperative control method is verified by experiments.The experimental results show that,compared with the traditional fixed trajectory tracking control,the model predictive control considering the uncertainty of human motion can reduce the maximum value and the root mean square value of human-robot interaction force,and can improve the wearing comfort.The risk sensitivity optimal control can adjust the tendency of the exoskeleton whether to track the reference trajectory or the actual movement trajectory of the wearer,and play the effect of risk seeking or risk avoidance.The variable factor risk sensitivity optimal control considering motion uncertainty can actively adjust the position of zero moment point and enhance the stability of the system when the system stability margin is low,and adjust the tracking performance of the load mass block and reduce the man-machine interaction force,which improves the wearing comfort when the stability margin is high.