Publication

Abstract : The application of meta-heuristic algorithms has significantly increased in recent years to find optimal solutions for continuous optimization problems. The Golden Jackal Optimizer (GJO) is a recently proposed swarm-based algorithm that has been considered to be a promising model of a meta-heuristic. Despite its superior performance, the GJO algorithm has flaws, including getting stuck in the local optimal regions and a lack of exploration ability, resulting in premature convergence when dealing with highly challenging optimization problems. Therefore, to circumvent this drawback, this paper proposes a Q-learning strategy combined with a novel adaptive mix-weighted dynamic opposition-based strategy (AMD) and random opposition-based learning (ROBL) strategy named the AMDRO-GJO algorithm. At first, the Q-learning method establishes a switching mechanism between AMD and ROBL strategies for the algorithm’s exploration. Lastly, during the updating phase, AMDRO-GJO identifies the best scheme for the global best solution, enhancing the algorithm’s exploitation. The effort of the proposed AMDRO-GJO algorithm has been examined on 23 classical, CEC2017, and CEC2019 benchmark functions. In addition, non-parametric tests such as the Wilcoxon rank sum test and t-test have been carried out to check the significance difference of the algorithms. Furthermore, the efficiency of several real-world engineering challenges has been evaluated by comparing it with those of rival optimizers. These experimental outcomes reveal the proposed AMDRO-GJO algorithm’s outstanding performance in tackling multiple optimization problems.