Redundancy resolution of a mobile manipulator using the KSOM based learning algorithm

Abstract

A learning-based strategy for the trajectory tracking of redundant mobile manipulators (MM) was presented in this study. A five-degrees-of-freedom (DOF) manipulator is mounted on the differential drive (DD) mobile robot. The advantage of using a redundant system is to avoid joint limits, obstacles, and singularities towards desired trajectory tracking. The proposed approach is based on the Kohonen Self-Organizing Map (KSOM) advanced with Weighted Least Norm (WLN) matrix algorithm. This approach is the recommended neural network for inverse kinematics solutions because of its stability, preserved topology, and capacity to optimize the joint space trajectory while producing a smooth minimal joint angle. A proposed method for redundancy resolution in MM has been simulated using MATLAB simulation code and the Gazebo real-time simulation physical environment. The simulation results are evaluated with the joint limit method of redundancy resolution and other existing controllers for verification purposes. The conventional method of redundancy resolution is local optimum and infeasible for the end-effector motion in the entire workspace. The KSOM uses different steps of error correction that improve the system’s performance as well as ensure the global asymptotical stability of the system. The Root Mean Square Error (RMSE) values for straight-line, circular, Lissajious, and irregular sinusoidal path motions of the proposed method using KSOM are given as 0.0095 m, 0.009945 m, 0.009897 m, and 0.009758 m, respectively. The simulation results of the proposed method confirm the effectiveness of the proposed approach.