Control of a 3DOF Robot Manipulator Using Inverse Jacobian and Multilayer Perceptron

This paper discusses the use of the inverse Jacobian method combined with neural network controller to improve statistical performance on a 3DOF robot manipulator. When using inverse Jacobian control, the kinematics of the robot manipulator requires proportional control which in this study is proposed to be replaced with multilayer perceptron type artificial neural network control. Based on the simulation results using MATLAB and obtained statistical data, i.e. standard deviations obtained for each end effector trajectory on the x, y and z axes are decreased from 11.25 to 9.679, from 14.27 to 13.93 and from 17.13 to 16.84, respectively. The reduction in standard deviation from experimental data shows that the robot arm movement trajectory with the proposed method is more homogeneous, consistent, and stable than the classical inverse Jacobian method.