Author

Yi Mei

View on Semantic Scholar

2 papers indexed here

We haven’t gathered this author’s papers yet. Follow them and we’ll fetch their work.

Not the right person? Other researchers publish under this name.

Open access Aug 2026

Cooperative co-evolution with adaptive decomposition schemes for large scale capacitated electric vehicle routing problem

The capacitated electric vehicle routing problem is a complex optimization problem consisting of two types of decisions: (1) deciding the routes for the electric vehicle fleets to complete the customer service; and (2) determining when to visit the charging stations. When the problem size grows, the problem becomes particularly challenging due to the large search space. To address this issue, we develop a cooperative co-evolution algorithm, containing novel problem decomposition and charging scheduling strategies. Two adaptive schemes are adopted to divide the problem into a reasonable number of subproblems with tractable sizes and select the customers with closeness beyond the threshold degree between two routes to measure their relationship. Based on the problem decomposition strategy, a memetic algorithm is designed as an optimizer for each subproblem, and the best sub-solutions are concatenated into an entire solution. The proposed algorithm is verified by comparing it with a number of state-of-the-art algorithms on two popular benchmark datasets as well as their enhanced forms. The experimental results show that our proposed algorithm outperforms the compared algorithms on nearly all instances. In particular, it successfully updated the majority of best-known solutions to the large instances.

Yuzhou Zhang, Yi Mei, Wen-Jie Xiao et al. · Memetic Computing · 0 citations

Preprint Jul 2026

LaT: LLM-as-Trainer for Multi-Task Vehicle Routing Solvers

Multi-task neural solvers aim to handle multiple Vehicle Routing Problem (VRP) variants within a unified model, avoiding separate training for each constraint combination. However, VRP variants differ in optimization difficulty, while existing methods lack stage-wise feedback on their training status, making the model biased to some specific variants. Although meta-learning can support adaptive training, it typically requires bi-level optimization and additional gradient updates, increasing computational cost. To address this limitation, we propose LLM-as-Trainer (LaT), a plug-and-play training paradigm that uses a pretrained large language model as an external trainer. LaT periodically analyzes cross-task validation metrics to generate a stage-wise guidance vector. This vector is combined with the current task's constraint vector and injected into each encoder layer, providing the neural solver with additional training information during subsequent policy optimization. Experiments on 16 VRP variants show that LaT improves the solution quality of several state-of-the-art multi-task neural solvers on both trained and unseen variants, supporting the effectiveness and generality of the proposed training paradigm.

Yang Wang, Yancong Jia, Wei-neng Chen et al. · 0 citations