Neural Network Bayesian Regularization Backpropagation to Solve Inverse Kinematics on Planar Manipulator

Abstract

Inverse kinematics is a behavior to find set joint angle value of the planar manipulator to reach end desire effector position. In this paper Bayesian regularization backpropagation training function is used to train neural network to produce the set of joint angle value to reach the desired position. Different architecture of the network also being tested to solve the inverse kinematics solution. A trainer planar manipulator used as a testbed of the proposed method. The robot performs nodes resembles square and triangle shape in its workspace based on neural network solution. The result shows the validity of the neural network solution to solve inverse kinematics of the planar manipulator.