The Study Of Varying-Parameter Recurrent Neural Network Based Redundant Robot Manipulator Self-motion Planning Scheme

In recent years,more and more robots are applied in our life.The robot manipulator with strong operation ability is a hot spot in the field of robot research.Redundant manipulator with more degrees of freedom is favored by researchers.The most important characteristic of the redundant manipulator is that it can carry out self-motion.Self-motion is that the manipulator's configuration moves to any desired state that satisfies the condition of self-motion from an initial state without moving the end-effector.Through self-motion,the robot manipulator could avoid obstacles,avoid physical limits of joints,solve the problem of joint-drift in repetitive motion,and some special tasks where the manipulator's end-effector is locked.This paper uses the idea of optimization and neural network solver to solve the optimal solution of the self-motion problem.Firstly,this paper introduces the significance of the research and the current research situation at home and abroad,and then introduces in detail the translation,rotation,composite transformation of coordinates,and its homogeneous transformation matrix.Finally,the D-H parameter method is used to establish the kinematics model of the manipulator.Secondly,a varying gain neural single-criterion self-motion planning scheme is proposed by using the ideas of quadratic programming and recurrent neural network.Firstly,the self-motion problem of the manipulator is transformed into a quadratic programming problem constrained by the kinematics equation.Then the self-motion trajectory of the manipulator is obtained by using the varying parameter recurrent neural network with fast convergence speed and high precision to solve the quadratic programming problem.This method takes the velocity two-norm as the optimization index to minimize the energy consumption of the whole system.Then,aiming at the shortcoming of varying gain neural single-criteria self-motion planning scheme,a varying gain neural bi-criteria self-motion planning scheme is proposed.This scheme takes the velocity two-norm and the velocity infinite-norm as the optimization function to effectively optimizes the energy consumption and the velocity peak of the manipulator,and makes the motion smoother.At the same time,the varying gain recurrent neural network with penalty function is used to solve the joint trajectory of the manipulator.The experimental results show that the proposed scheme can effectively overcome the joint limit and velocity spike problems.Lastly,the self-motion scheme proposed in chapter four is extended to the mobile robot manipulator.With the addition of the mobile platform,the mobile manipulator increases the redundancy of the system,which increases the difficulty of system planning and reduces the accuracy.Experimental results show that the proposed scheme has good performance on the mobile manipulator.