OD-DEAL: Dynamic Expert-Guided Adversarial Learning with Online Decomposition for Scalable Capacitated Vehicle Routing

Dongbin Jiao, Zisheng Chen, Xianyi Wang, Jintao Shi, Shengcai Liu, Shi Yan · 2026

Read OD-DEAL: Dynamic Expert-Guided Adversarial Learning with Online Decomposition for Scalable Capacitated Vehicle Routing by Dongbin Jiao, Zisheng Chen, Xianyi Wang, Jintao Shi, Shengcai Liu, Shi Yan (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-00488.txt.

Description

Dongbin Jiao and colleagues address large capacitated vehicle routing, where classic heuristics grow complex and neural solvers generalise poorly. Their OD-Gear scheme joins hybrid genetic search with online barycentre clustering and group-relative optimisation, distilling expert heuristics into a graph attention network policy. Adversarial training turns divide-and-conquer tactics into dense reward signals, while group sampling uses relative gains to protect both diversity and quality. Inference needs no clustering step, runs fast, and stays near the front at the 10,000-node scale, giving heuristic-quality routes for massive graphs.

Catalog record

CREATORDongbin Jiao, Zisheng Chen, Xianyi Wang, Jintao Shi, Shengcai Liu, Shi Yan
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE112646
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602.00488

Browse the full library · About the Tower of Babel