Analysis of the inverse kinematics of redundant manipulators is one of the nesseccary tools in various robotic fields such as design, motion planning and control of these systems. Since, there is not an analytical solution for the inverse kinematics of several redundant manipulators, numerical approaches are needed to execute and investigate in this field. In this paper, combination of the neural networks, fuzzy systems and quadratic programming is used to obtain the joint variables. According to the proposed approach, seven neural networks are considered according to the each joint variable and by adaptation of the neural network’s weights, suitable configurations of the robot is determined to track a desired trajectory in the Cartesian space. Meanwhile, initial weights of the neural networks are obtained by fuzzy systems based on the vicinity of the end-effector to desired point and feasibility of the joint variables. Obstacle avoidance scheme is performed by investigation of the conditions including choosing the joint variables that involved in the equations of the arms constraints and determination of the most critical arm. In order to establish the constraints of the problem in the quadratic programming, realization of the Kun-Tucker conditions will be used. Evaluation of the proposed approach will be carried out on the PA-10 manipulator by simulation and analysis of the results.

Keywords: Inverse Kinematics, Redundant Manipulator, Neural Network, Quadratic Progarmming, Fuzzy System, Obstacle Avoidance.

Full-Text [PDF 1648 kb]   (4003 Downloads)