Adaptive Repetitive Motion Planning For Redundant Manipulator At Velocity Level

Manipulator inverse kinematics planning is an important research direction in mechanical industry.For the manipulator inverse kinematics planning,the traditional pseudo-inverse solution cannot guarantee that the manipulator does not exceed the physical constraints.The traditional quadratic programming solution needs to know the accurate physical parameters of the manipulator.In practical applications,it often occurs that the physical parameters of the manipulator caused by physical wear are different from the real value,and the joint of the manipulator exceeds the physical constraints,which leads to the damage of the manipulator.In the above cases,the traditional solutions is not applicable.An inverse kinematics solution based on adaptive neural network was designed by discussing the manipulator inverse kinematics planning in this paper.For manipulators with unknown or wrong physical parameters,this solution can ensure that the joint of the manipulator does not exceed the physical limit to complete repeated motion.The main work and achievements of this paper are as follows:1.Aiming at the problem of repetitive motion of manipulator,a solution of repetitive motion based on quadratic programming was discussed.By minimizing the difference between the current value of joint angle and the initial value of joint angle of manipulator,andcombining with neurodynamic method,a performance index which could realize repetitive motion was designed,which make the joint angle return to the initial position after the end of the motion.Then,the Lagrangian method was used to transform the quadratic programming into a linear system to solve the problem.The experimental results verify the effectiveness of the performance index to achieve repetitive motion.2.Aiming at the trajectory planning problem of the manipulator with unknown model,an adaptive trajectory planning solution based on pseudo-inverse was discussed.The suitable Jacobian matrix of the manipulator is found by the zeroing neural network,which gets rid of the dependence on the mathematical model of the manipulator.The unknown manipulator model can be controlled only by input and output information.Finally,the feasibility of the solution to deal with unknown models was verified by theoretical analysis and experimental results.3.In order to make the joint angle of the manipulator with unknown model not exceed its physical constraint on the basis of repeated motion,an inverse kinematics solution based on adaptive neural network was proposed.Aiming at the physical constraint of the manipulator,the joint angle constraint was introduced,and a projection neural network was used to solve the quadratic programming with physical constraints.In order to deal with the unknown parameters,the Jacobian matrix of the manipulator was decomposed into the parts related to the unknownparameters.And an adaptive neural network was designed combining with the projection neural network.The feasibility of the solution was proved by theoretical analysis,and the superiority of the solution was fully verified by the experimental results.