Feature Scaling for Multi-objective Optimization Applied to Solving the Inverse Kinematics of Redundant Robot Manipulators

This study proposes the implementation of feature scaling in the definition of a multi-objective function to solve the inverse kinematics problem of redundant robot manipulators. The case study consisted of generating a desired kinematic path for an 8-degree-of-freedom (DOF) robotic system with a specific position and orientation. A multi-objective function is defined with feature scaling to ensure that the position and orientation errors operate on the same scale. Particle swarm optimization (PSO) and differential evolution (DE) were used to solve the optimization problem and compare the performance of the defined objective function with feature scaling against a weight-based function. The results show that the proposed multi-objective function with feature scaling improves the performance of PSO and DE in terms of the execution time, number of iterations and accuracy in finding optimal solutions.