An Adaptive Hybrid Genetic and Large Neighborhood Search Approach for Multi-Attribute Vehicle Routing Problems

Weiting Liu,Yunqi Luo,Yugang Yu

2024-02-29

Abstract:Known for its dynamic utilization of destroy and repair operators, the Adaptive Large Neighborhood Search (ALNS) seeks to unearth high-quality solutions and has thus gained widespread acceptance as a meta-heuristic tool for tackling complex Combinatorial Optimization Problems (COPs). However, challenges arise when applying uniform parameters and acceptance criteria to diverse instances of the same COP, resulting in inconsistent performance outcomes. To address this inherent limitation, we propose the Adaptive Hybrid Genetic Search and Large Neighborhood Search (AHGSLNS), a novel approach designed to adapt ALNS parameters and acceptance criteria to the specific nuances of distinct COP instances. Our evaluation focuses on the Multi-Attribute Vehicle Routing Problem, a classical COP prevalent in real-world semi-automated storage and retrieval robotics systems. Empirical findings showcase that AHGSLNS not only competes effectively with ALNS under varying parameters but also exhibits superior performance in terms of convergence and stability. In alignment with our dedication to research transparency, the implementation of the proposed approach will be made publicly available.

Optimization and Control

What problem does this paper attempt to address?

This paper mainly discusses the Multi-Attribute Vehicle Routing Problem (MA VRP), which is a classic problem in combinatorial optimization and is commonly seen in the task scheduling of automated storage and retrieval systems. The authors propose a new algorithm called Adaptive Hybrid Genetic Search and Large Neighborhood Search (AHGSLNS) to address the problem of inconsistent performance of complex optimization problems for different instances when applying uniform parameters and accepting criteria.

The traditional Large Neighborhood Search (LNS) algorithm seeks high-quality solutions through the strategy of disruption and repair, but its performance is influenced by parameter selection and acceptance criteria, and may have different effects for different problem instances. Therefore, the paper proposes AHGSLNS, which combines LNS with Genetic Search (GS), and adapts parameter and acceptance criteria through parallel mechanisms to improve stability and performance in different problem instances.

AHGSLNS is validated on the Multi-Attribute Vehicle Routing Problem, which includes multiple warehouses, vehicles, customers, and takes into account various situations such as return trips and open routes. By mathematically modeling the problem and using integer programming methods, the paper demonstrates how AHGSLNS improves algorithm convergence and stability through adaptive survival mechanism and Cooperative Evolution Mechanism (CEM) while ensuring solution quality.

Experimental results show that AHGSLNS can achieve optimal solutions in small-scale instances and has significant improvements in convergence speed and stability compared to LNS in large-scale instances. Furthermore, compared to various variants of LNS algorithms, AHGSLNS exhibits better performance in most cases.

In summary, the goal of this paper is to address the performance differences caused by unified parameters and acceptance criteria in the Multi-Attribute Vehicle Routing Problem. Through proposing the AHGSLNS algorithm, adaptive optimization for different problem instances is achieved, resulting in improved efficiency of solution and solution quality.