Adaptive Evolutionary Multi-Tasking for Solving Two-Echelon Routing Problem
Nannan Zuo, Rong Hu, Qingxia Shang, Bin Qian · 2025
The two-echelon vehicle routing problem (2E-VRP), as a variant of the VRP, is an NP-hard problem. Traditional methods for solving the 2E-VRP typically focus on independently optimizing individual problems using mathematical approaches or evolutionary algorithms. However, these methods are prone to getting stuck in local optima. In light of this, a novel adaptive evolutionary multi-tasking(AEMT) algorithm is proposed in this paper to solve the 2E-VRP. Specifically, the algorithm first constructs a multi-task framework based on the 2E-VRP. The two-echelon transportation network in the 2E-VRP is decomposed into multiple subproblems at satellites, with each secondary subproblem being treated as a separate task. Next, the optimization performance of each task is enhanced through knowledge sharing between tasks. To avoid the emergence of shorter sub-routes, an adaptive multi-task strategy is employed, which dynamically adjusts the relationship between customers and satellites. Tasks are dynamically generated as the evolution progresses. To evaluate the effectiveness of the MTEA algorithm, standard test instances of the 2E-VRP are used, and the proposed algorithm's effectiveness is verified through simulation experiments and algorithm comparisons.