<![CDATA[There are many new metaheuristic algorithms (MA) introduced in recent years. Most of them employ multiple search strategy. Unfortunately, many of these algorithms accommodate this multiple search strategy by using predefined mechanism which is not adaptive on handling circumstance during optimization process. Moreover, many of these mechanisms are memoryless. Due to this problem, we propose a new decision-making model to handle this multiple search strategy developed based on reward-based mechanism which is adopted from reinforcement learning (RL). Then, this mechanism is implemented into a new existing MA known as glider snake optimization (GSO) so that it becomes variant of it known as reward-based glider snake optimization (RGSO). Two variants are introduced in this work which are RGSO-1 and RGSO-2. RGSO-2 is an extended version of RGSO-1 as an additional search is included. The performance of both variants is investigated using 23 classic functions and three cases of economic load dispatch problems (ELDP). Five existing algorithms including GSO are chosen as benchmarks. The result shows that RGSO-2 is superior to GSO and all other benchmarks in almost all functions and maintains its competitiveness in ELDP. RGSO-2 is better than GSO in 12 functions. Meanwhile, RGSO-1 is comparable to GSO in all functions and ELDP. Moreover, RGSO-1 achieves the optimal solution for two functions while RGSO-2 achieves the optimal solution for five functions.]]>
Copyrights © 2026
