Answers are grounded in this paper and cite their sources.
In this paper, a new mathematical model and a novel meta-heuristic are presented. Unlike classic routing problems, the proposed model considers drivers’ preferences by preventing scattered customers in routes. To address a real-world case, a Learning-based Adaptive Large Neighbourhood Search (ALNS) is introduced. The proposed approach employs Q-learning as a specific Reinforcement Learning method for operator selection. This learning-based approach outperforms the traditional roulette wheel and random selection method. Moreover, four new operators are introduced with the ability to improve the objective function in the last quarter iterations. The proposed algorithm was compared with classic ALNS and VNS using 30 test problems in different scales. The results indicate that the objective function reached by the proposed algorithm is on average 3.51% and 19.34% better than the objective function of VNS and classic ALNS, respectively. Also, the Q-learning approach outperforms the roulette wheel and random selection by 0.69% and 0.66%, respectively.
No takes yet. Share an insight, caveat, or question.
Goodarzi et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: