A nested optimization approach for robot gripper multi-objective optimization problem

Abstract

Robot Gripper Design Optimization Problem is a well-known multi-objective optimization problem explored by various approaches in previous studies over the last few decades. The continuous parameter z plays an important role in determining a robot gripper's minimum and maximum gripping forces. However, previous studies have utilized a simple loop approach in which the value of z is iteratively increased by a constant value from 0 to 50 to determine the gripping forces of a solution. This approach limits the solution space by treating the continuous z parameter as a discrete variable. To overcome this limitation, we propose a nested optimization approach in which a single-objective optimizer is embedded within the fitness function of the main optimizer to determine the optimal value of z. The proposed approach is evaluated using a single-optimizer Simulated Annealing (SA) nested within multi-objective optimizers such as Multi-Objective Gray Wolf Optimizer (MOGWO), Multi-Objective Particle Swarm Optimization (MOPSO), Multi-Objective Artificial Algae Algorithm (MOAAA), and Non-Dominated Sorting Genetic Algorithm (NSGA-II). Experimental results demonstrate that our approach significantly outperforms previous studies and achieves the best-known performance. The findings of the study indicate that this method can present a novel framework for researchers in tackling design optimization challenges, encouraging a reevaluation and possible improvement of conventional techniques.