Machine Learning approach for Inverse Kinematics in Trajectory Planning of Pioneer 2 Manipulator with Cubic Spline Interpolation

Abstract

The primary objective of robot manipulators is to achieve the desired orientation and point of end effector in order to accomplish the pre-established task. Inverse kinematic analysis will be used in the pioneer 2 robot to obtain a successful solution to design and operate the arm. This paper considers a 5-dof revolute Pioneer2 manipulator which is compact, low cost and lightweight. When the DOF of the robot increases, the inverse kinematic problem becomes more and more complex and gives n number of joint configurations for the same position. This results in making the standard solution for this problem becomes trickier. To overcome the computational complexity of kinematic analysis of Pioneer 2 robot, the objective of this study is to perform intelligent computation of inverse kinematics with the use of machine learning techniques that consists of linear regression, K-Nearest Neighbor algorithm and Artificial Neural Network. By comparing three algorithms R-square values and RMSE values, it is observed that KNN algorithm is giving better results. Therefore, KNN can be used for better solution of inverse kinematics with fast results and high accuracy. Then the smooth trajectory is achieved using cubic spline interpolation.