Adaptive Parameter Setting for Genetic Algorithms Using Reinforcement Learning: A Case Study on the Capacitated Vehicle Routing Problem

作者：Vipul Razdan · 发表于：2026 3rd International Conference on Emerging Trends in Engineering and Medical Sciences (ICETEMS) · 年份：2026 · DOI：10.1109/icetems66917.2026.11469619 · 被引用次数：2

This paper uses a combination of Reinforcement Learning and Genetic Algorithm to adaptively optimize parameters in solving the Capacitated Vehicle Routing Problem through the use of Node Representation. When using a traditional static approach like Design of Experiments (DOE), the diversity of the population pool deteriorates, which does not yield the best solution. The RL-GA Approach Updates GA Parameters Dynamically Leading To Improved Quality Of Solutions On A Problem Set Of Benchmark CVRP Problems. The RL-GA always performs better than static methods and has the potential for greater application in other categorical and non-linear problems. In the future, the RL-GA will be tested on larger CVRP instances, which will speed up the quest for a solution to complex practical problems using genetic algorithms.