PASA: A Priori Adaptive Splitting Algorithm for the Split Delivery Vehicle Routing Problem

TL;DR

This paper introduces PASA, an adaptive splitting algorithm for the split delivery vehicle routing problem that improves solution speed and quality by considering customer demand and location, outperforming fixed split strategies.

Contribution

The paper proposes an adaptive a priori splitting rule for SDVRP that enhances solution efficiency and effectiveness over existing fixed splitting methods.

Findings

PASA generates solutions comparable to state-of-the-art methods.

PASA outperforms fixed a priori splitting rules in experiments.

The adaptive rule reduces computational time significantly.

Abstract

The split delivery vehicle routing problem (SDVRP) is a relaxed variant of the capacitated vehicle routing problem (CVRP) where the restriction that each customer is visited precisely once is removed. Compared with CVRP, the SDVRP allows a reduction in the cost of the routes traveled by vehicles. The exact methods to solve the SDVRP are computationally expensive. Moreover, the complexity and difficult implementation of the state-of-the-art heuristic approaches hinder their application in real-life scenarios of the SDVRP. In this paper, we propose an easily understandable and effective approach to solve the SDVPR based on an a priori adaptive splitting algorithm (PASA). The idea of a priori split strategy was first introduced in Chen et al. (2017). In this approach, the demand of the customers is split into smaller values using a fixed splitting rule in advance. Consequently, the original SDVRP instance is converted to a CVRP instance which is solved using an existing CVRP solver. While the proposed a priori splitting rule in Chen et al. (2017) is fixed for all customers regardless of their demand and location, we suggest an adaptive splitting rule that takes into account the distance of the customers to the depot and their demand values. Our experiments show that PASA can generate solutions comparable to the state-of-the-art but much faster. Furthermore, our algorithm outperforms the fixed a priori splitting rule proposed by Chen et al. (2017).