Solving the Green Vehicle Routing Problem with Simultaneous Pickups and Deliveries via Instance Reduction and Variable-Depth Neighborhood Search

This paper presents a Variable-Depth Neighborhood Search approach for solving the Green Vehicle Routing Problem with Simultaneous Pickup and Delivery. The method explores reduced subinstances modeled as Integer Linear Programs and solved by an exact solver. Strategies are incorporated to guide the neighborhood definition, based on customer proximity and the reuse of multiple previously found solutions. These mechanisms foster a balance between intensification and diversification. Computational experiments on benchmark instances demonstrate that the proposed approach outperforms existing methods in both solution quality and robustness, particularly on large-scale problems. An ablation study confirms the positive impact of the adopted strategies on overall performance.