Towards Generalization-Oriented Models for Vehicle Routing Problems with Mixture-of-Experts
About
In recent years, Deep Reinforcement Learning (DRL) has achieved substantial progress on Vehicle Routing Problems (VRPs). However, existing DRL-based methods are typically trained on instances generated from a uniform distribution, which limits their performance under real-world distribution shifts. In this paper, we aim to develop a generalization-oriented model that partitions the policy network into multiple modules and adaptively recombines modules to form specific policies during inference. Specifically, we propose Residual Refined Experts with Instance-level Gating (R2E-IG) to improve cross-distribution generalization. Our contributions are threefold: (1) We introduce a Residual Refined Expert (R2E) architecture that enhance expert expressiveness via residual refinement; (2) We design an instance-level gating mechanism that learns distribution-aware instance representations and routes inputs to suitable modules; (3) We propose a mixed-distribution training mechanism equipped with Dynamic Weight Adaption (DWA), which dynamically reweights training data from different distributions to emphasize more informative ones. Extensive experiments show that R2E-IG achieves competitive performance against state-of-the-art baselines on both in-distribution and out-of-distribution instances across synthetic and benchmark datasets. Moreover, R2E-IG is generic and can be easily integrated into existing DRL-based methods to further improve performance.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Capacitated Vehicle Routing Problem | CVRP N=100 | Objective Value15.6268 | 87 | |
| Capacitated Vehicle Routing Problem | CVRP 20 | Objective Value6.0801 | 43 | |
| Capacitated Vehicle Routing Problem | Expansion CVRP20 | Objective Value5.3767 | 18 | |
| Capacitated Vehicle Routing Problem | Explosion CVRP100 | Objective Value12.3139 | 18 | |
| Capacitated Vehicle Routing Problem | Expansion CVRP50 | Objective Value8.2284 | 9 | |
| Capacitated Vehicle Routing Problem | Expansion CVRP100 | Objective Value11.4175 | 9 | |
| Capacitated Vehicle Routing Problem | Explosion CVRP50 | Objective Value8.8002 | 9 | |
| Capacitated Vehicle Routing Problem | Grid CVRP50 | Objective Value10.4649 | 9 | |
| Capacitated Vehicle Routing Problem | Implosion CVRP50 | Objective Value10.3022 | 9 | |
| Traveling Salesman Problem | Expansion TSP 50 | Objective Value4.3897 | 9 |