Extraction And Application Of Human Upper Limb Motion Control Strategy

With the acceleration of global population aging and the increase of labor cost,robots are widely demanded and used in practical industrial fields.As the extension of real-word applications,robots are not only constrained in industrial fields,but also widen to human daily life.Compared with the traditional industrial fields,the dynamic characteristics of human daily living environment are more complicated,which leads to the traditional trajectory planning control method based on pre-programming is inadequate to handle daily-life task requirements.Thus,the issues of investigating the control strategies involved in human motions and transferring the obtained control strategies to robot,such that the robot can obtain human-like control behavior,are the keys to enhance the working ability of robot in dynamic environment.It is well known that human control system is composed by the upper central nervous system and the lower muscle reflex control.However,due to the uncertainties and complexities of human physiological structure,the control strategies behind different human motions are unclear,which results in revealing the control strategies via direct construction approaches to be difficult.To this end,this thesis devotes its research interest on studying the construction and extraction methods of human control strategy,as well as the mapping method regarding to human control strategy into robot system.The main works of this thesis are summarized as follows:(1)Construction and extraction of human motion control model based on bi-level optimization.Firstly,the inverse optimization model of human upper limb motion control behavior is constructed.Then,a bi-level optimization framework based on particle swarm optimization(PSO)and direct collocation(DC)is proposed to solve the established motion control model and extract the control strategy.(2)Learning human motion control strategy based on model predictive control(MPC).Firstly,the MPC prediction model of human upper limb motion control is established.Next,a PSO-based indirect mapping method is designed to transmit the objective function of human motion control strategy to that of the MPC prediction model.Lastly,the MPC predictive optimal control law is solved by adding a feedback correction strategy to the established MPC control model.(3)Experimental studies and results analysis.The efficacy of the developed bi-level optimization method is verified by two typical human upper limb motions,i.e.,hitting ball and moving object.Moreover,the effectiveness of the proposed control mapping method is tested by the experimental tests conducted these two typical human motions on UR5 programmed under ROS environment.The experimental results reveal that the developed bi-level optimization method can effectively extract the control strategies of the two aforementioned human upper limb motions.Also,the MPC based control framework can effectively transfer the human control strategy to the UR5 manipulator.Moreover,compared with the direct mapping method,the indirect mapping method can obtain the control strategy more suitable for the robot ontology.Thus,the proposed methods proposed could provide a reference for the construction of robot human control research.