Szczegóły publikacji
Opis bibliograficzny
Island-DR-ALNS: a diversity-driven multi-agent framework for deep reinforcement learning-based combinatorial optimization / Adam Żychowski, Xin Yao, Jacek MAŃDZIUK // W: Parallel Problem Solving from Nature – PPSN XIX : 19th international conference : PPSN 2026 : Trento, Italy, August 29–September 2, 2026 : proceedings , Pt. 5 / eds. Giovanni Iacca, [et al.]. — Cham : Springer Nature, cop. 2027. — ( Lecture Notes in Computer Science ; ISSN 0302-9743 ; vol. 16989 ). — ISBN: 978-3-032-36216-2; e-ISBN: 978-3-032-36217-9. — S. 70–86. — Bibliogr., Abstr. — Publikacja dostępna online od: 2026-08-25. — J. Mańdziuk - dod. afiliacja: Warsaw University of Technology, Warsaw, Poland
Autorzy (3)
- Żychowski Adam
- Yao Xin
- AGHMańdziuk Jacek
Dane bibliometryczne
| ID BaDAP | 170356 |
|---|---|
| Data dodania do BaDAP | 2026-10-29 |
| DOI | 10.1007/978-3-032-36217-9_5 |
| Rok publikacji | 2027 |
| Typ publikacji | materiały konferencyjne (aut.) |
| Otwarty dostęp |  |
| Wydawca | Springer |
| Konferencja | Parallel Problem Solving from Nature 2026 |
| Czasopismo/seria | Lecture Notes in Computer Science |
Abstract
Reinforcement Learning-based Combinatorial Optimization (RLCO) has emerged as a powerful alternative to handcrafted heuristics in CO. Most existing RLCO approaches rely on a monolithic learning agent that is prone to premature convergence and lack of behavioral diversity. To address this, we propose Island-DR-ALNS, a novel framework that integrates Deep Reinforcement Learning with a decentralized island model architecture. Instead of a single agent, the proposed method deploys a population of heterogeneous agents that cooperatively optimize the parameters of the Adaptive Large Neighborhood Search (ALNS). To increase agents diversity, Negative Correlated Search is incorporated as a regularization mechanism, which explicitly rewards agents for developing distinct search policies. Furthermore, we introduce a socially-aware state representation that enables agents to contextualize their performance relative to others. Extensive experiments on the Capacitated Vehicle Routing Problem demonstrate that Island-DR-ALNS outperforms state-of-the-art baselines. The framework also exhibits strong generalization capabilities, enabling effective zero-shot transfer to distinct problems such as TSP and mTSP without retraining.
