Efficient Constrained Real-World Problem-Solving Using Gradient-Based Mutation Manta Ray Foraging Optimization (GM-MRFO)

Abstract

This paper presents a Gradient-based Mutation Manta ray foraging optimization (GM-MRFO) that is designed to solve real-world optimization problems with constraints. GM-MRFO combines the basic strategy of MRFO with the Gradient-based Mutation (GM) strategy, which is a feasibility-and-solution repair strategy adopted from the ϵ-Matrix-Adaptation Evolution Strategy ( MAgES). MRFO algorithm is not immune to the common problems confronted by constrained optimization algorithms where constraints in the optimization problem are incompatible, and a solution that satisfies all constraints does not exist. In such cases, the MRFO algorithm may not be able to find a feasible solution. Another challenge is the optimization algorithm converges to a solution that is not globally optimal. By introducing the GM strategy and using Jacobian approximation in finite differences, GM-MRFO can improve the feasibility of solutions throughout the search process, which enables it to handle constraints more effectively than its predecessor. The proposed algorithm's performance is evaluated by assessing the accuracy of the best solution produced, the feasibility rate, mean of violation, success rate, and the ranking based on a ranking scheme in the Congress on Evolutionary Computation 2020 (CEC2020). Specifically, GM-MRFO achieved a feasibility rate of 100% on 47 out of 57 CEC2020 real-world problems and improved adequately best-known solutions.