Research On The Perception And Control Method Of Human Robot Cooperation Based On Machine Learning

With the rapid development of robotics technology,robots have gradually changed from simple tasks that are independently completed to human-robot cooperation to complete more complex operations.Therefore,human-robot cooperation technology has broad application prospects and important research value.In the process of human-robot cooperation,it is expected that people and robots can operate safely and comfortably.The purpose of this study is to improve the stability and security of human-robot cooperation systems,and to make human-robot collaboration more flexible and comfortable.Firstly,human-robot cooperation model is analyzed,and the equivalent model of human-robot cooperation is established.the human upper limb model is simplified,the kinematics and dynamics equations of the arm are established by D-H method and Lagrangian mechanics analysis method respectively.The human-robot interaction force information is used to infer the operator's movement intention,and be made as the reference input information for the controller to control human-robot coordination movement.Secondly,using the human-robot cooperation model to determine the coefficients in the impedance control experiment,which provides data support for the following research based on impedance control.the control system uses a two-dimensional fuzzy controller to realize the adjustment of interactive force information in the contact human-robot cooperation process.The experimental data of the robot's terminal velocity and interaction force during the human-robot cooperation movement are analyzed,and compared with the results of the impedance control algorithm,the superiority of the fuzzy control method over the impedance control method is verified.Then,in order to further improve the effect of human-robot cooperation control,through the judgment and analysis of the collaborators' behavior information,a method based on machine learning to identify collaborators' motion intentions was proposed.A BP neural network model was established to train robot's intent recognition training so that it has a certain ability of learning and forecasting,so that it can realize online estimation of human intention according to the force of the partner and the movement characteristics of the robot during the cooperation process.The advantage of this method lies in overcoming the shortcomings of complex and varied human motion models in traditional methods,difficulties in establishing human-robot cooperation models,and difficult to estimate human body impedance parameters.At last,a human-robot cooperation system with one degree of freedom is taken as a platform,and the background is studied by human-robot cooperation.The content covers the contact information estimation of robot joint arm and the cooperation intention identification.It changed the traditional idea of adjusting the compliant control parameters to realize the flexibility of the robot and expanded the research field of the compliant control method.The results show that the control method proposed in this paper can better achieve the coordinated operation of human-robot to ensure the stability,safety and flexibility of the system.