摘要
The Traveling Salesman Problem (TSP) is a classical NP-hard combinatorial optimization problem with many practical variants. The Lin-Kernighan-Helsgaun (LKH) algorithm is one of the state-of-the-art local search algorithms for the TSP. LKH-3 is a powerful extension of LKH that can solve many TSP variants. Both LKH and LKH-3 use the candidate sets to improve their efficiency, and have two different methods, alpha-measure and POPMUSIC, to decide the candidate sets. In this work, we first propose a Variable Strategy Reinforced LKH (VSR-LKH) algorithm, which incorporates three reinforcement learning methods (Q-learning, Sarsa, and Monte Carlo) with LKH, for the TSP. We further propose a new algorithm called VSR-LKH-3 that combines the variable strategy reinforcement learning method with LKH-3 for typical TSP variants, including the TSP with Time Windows (TSPTW) and the Colored TSP (CTSP). The proposed algorithms replace the inflexible traversal operations in LKH and LKH-3 and let the algorithms learn to make a choice at each search step by reinforcement learning. Both LKH and LKH-3, with either the alpha-measure or the POPMUSIC method, can be significantly improved by our methods. Extensive experiments on 236 widely-used TSP benchmarks with up to 85,900 cities demonstrate the excellent performance of VSR-LKH. VSR-LKH-3 also significantly outperforms the state-of-the-art heuristics for the TSPTW and CTSP.
学科与关键词
- Reinforcement learning for the traveling salesman problem with refuelingOttoni, Andre L. C.*; Nepomuceno, Erivelton G.; Oliveira, Marcos S. de; Oliveira, Daniela C. R. deComplex & Intelligent Systems, 2022, 8(3): 2001-2015.被引 25
- Solving Dynamic Traveling Salesman Problems With Deep Reinforcement LearningZhang, Zizhen; Liu, Hong*; Zhou, MengChu; Wang, JiahaiIEEE Transactions on Neural Networks and Learning Systems, 2023, 34(4): 2119-2132.被引 50
- Dynamic sub-route-based self-adaptive beam search Q-learning algorithm for traveling salesman problemZhang, Jin*; Liu, Qing; Han, XiaoHangPLoS One, 2023, 18(3): e0283207.被引 2
