Enhancing automatically designed relocation rules with the rollout algorithm

Created by W.Langdon from gp-bibliography.bib Revision:1.8414

@Article{Durasevic:2025:swevo,

• author = "Marko Durasevic and Mateja Dumic and Francisco Javier {Gil Gala}",
• title = "Enhancing automatically designed relocation rules with the rollout algorithm",
• journal = "Swarm and Evolutionary Computation",
• year = "2025",
• volume = "96",
• pages = "101975",
• keywords = "genetic algorithms, genetic programming, Container relocation problem, Relocation rules, Rollout algorithm",
• ISSN = "2210-6502",
• URL = "https://www.sciencedirect.com/science/article/pii/S2210650225001336",
• DOI = "doi:10.1016/j.swevo.2025.101975",
• abstract = "The container relocation problem (CRP) is a complex optimisation problem in maritime transport. To solve this problem, heuristic approaches are often used, ranging from relocation rules (RRs) to metaheuristics. Although metaheuristics outperform RRs, the latter remain popular due to their simplicity and adaptability. The manual design of RRs is challenging, which is why genetic programming (GP) is used to automatically generate them. However, RRs generated by GP generally achieved inferior solutions compared to metaheuristics. To close this gap, this study applies the rollout method to improve the performance of RRs while maintaining reasonable execution times. The rollout algorithm strikes a balance between exhaustive and heuristic search by combining partial enumeration with RR-based decision evaluation. Although the rollout method improves the quality of the solution, it also leads to considerable computational cost. To solve this problem, three strategies for reducing the search space are proposed. Experimental results show that the rollout algorithm significantly improves solution quality compared to standard RRs, with the proposed search space reduction techniques effectively reducing execution time without compromising performance. In particular, the results show that the rollout algorithm can be executed 2 to 4 times faster using the proposed reduction techniques, while its performance is reduced only by 1percent",

}

Genetic Programming entries for Marko Durasevic Mateja Dumic Francisco Javier Gil Gala

Citations