Capsule aggregated attention for vehicle routing problems

Deep learning-based methods have shown great potential for solving the vehicle routing 问题 (VRP) in recent years. In the current learning-based models, attention mechanism plays an important role and becomes one of the key modules for improving the performance. However, the aggregate-by-summation paradigm of attention is not expressive enough to fully capture the rich information in the VRP. To solve this 问题, we propose a novel capsule aggregated attention mechanism, which utilizes capsule to store more information, applies dynamic routing for information aggregation, exploits soft gated capsule selector to differentiate the importance of different capsules, and modifies the context node vector in the decoding process to reflect the state changes. Based on the above modifications, we design a new graph attention network for solving the VRP under the reinforcement learning framework in this paper, and validates the effectiveness of our proposed method through sufficient numerical experiments.