Research Of Trajectory Tracking Control Method For Manipulator Based On Neural Network

With the rapid development of robot technology, industrial manipulators have been widely used in the civilian and military fields, especially playing an irreplaceable role in electronic, outer space and the nuclear industry. Therefore, in order to make the manipulator operate rapidly, accurately and efficiently, the research on reasonable trajectory planning and trajectory tracking control is particularly significant. This paper takes the PUMA560 manipulator with six degrees of freedom (DOF) as the research object and carries on in-depth research and analysis on the inverse method, trajectory planning method and trajectory tracking control strategy of manipulator.Firstly, this paper introduces some related knowledge of the rigid body posture and the coordinate transformation. According to the structural characteristics of PUMA560 manipulator, the Denavit-Hatenberg (D-H) method is adopted to establish the kinematics equation of manipulator. By using MATLAB to do the simulation experiment of kinematics, the experimental results confirm that the positive solution of the manipulator kinematics is unique and the inverse solutions are not unique. Simultaneously, this paper makes use of Lagrange method to establish the dynamic model of a serial manipulator.Secondly, in view of the reason that conventional methods for solving inverse kinematics have the defects of large amount of calculation, slow convergence rate and low accuracy. A parallel neural network method based on BP and RBF neural network for solving inverse kinematics of robot is presented in this paper. BP network adopts Levenberg-Marquardt (LM) algorithm suitable for processing large amounts of data and RBF network increases radial basis neurons automatically. Simulation results show that the parallel neural network algorithm for solving inverse kinematics of robot proposed in this paper is simple and credible, laying a solid theoretical basis for the trajectory planning and trajectory tracking control of the manipulator.Then, this paper detailed analyzes the method of manipulator trajectory planning in joint space. Using inverse kinematics to solve the joint variable values of the path points the manipulator passes through and fully considering the kinematics constraints, this paper proposes a smooth trajectory planning method using five B-spline curves to build joint trajectory. The simulation results show that this method ensures the velocity, acceleration and jerk of each joint trajectory are continuous and the velocity, acceleration and jerk of the initiation and termination point can be arbitrarily configured. This method provides a desired trajectory for trajectory tracking control experiments of the manipulator.Finally, considering that the traditional control algorithms are limited by the uncertainty of dynamic model and result in degradation of control performance, this paper proposes an RBF neural network robust control algorithm using RBF network to approximate the uncertainty of the dynamics model and restraining the approximation error of neural network by the robust control term. In the meantime, this paper designs a serial manipulator with three DOF and conducts a trajectory tracking control experiment making the manipulator move along the joint trajectory which has been planned. The experimental results show that the RBF neural network robust control algorithm presented in this paper has high control accuracy and strong robustness and verify the validity and feasibility of this algorithm.