[Paper Review] Heuristics for Vehicle Routing Problem: A Survey and Recent Advances

Fei Liu, Chengyu Lu+4 more|arXiv (Cornell University)|Mar 1, 2023

Vehicle Routing Optimization Methods17 citations

TL;DR

This paper systematically surveys vehicle routing heuristics (constructive, improvement, and metaheuristics), outlines a general framework, and discusses emerging topics like unified heuristics, automatic design, and machine learning assistance.

ABSTRACT

Vehicle routing is a well-known optimization research topic with significant practical importance. Among different approaches to solving vehicle routing, heuristics can produce a satisfactory solution at a reasonable computational cost. Consequently, much effort has been made in the past decades to develop vehicle routing heuristics. In this article, we systematically survey the existing vehicle routing heuristics, particularly on works carried out in recent years. A classification of vehicle routing heuristics is presented, followed by a review of their methodologies, recent developments, and applications. Moreover, we present a general framework of state-of-the-art methods and provide insights into their success. Finally, three emerging research topics with notable works and future directions are discussed.

Motivation & Objective

• Classify vehicle routing heuristics into constructive, improvement, and metaheuristics and summarize methodologies, developments, and applications.
• Provide a general framework for state-of-the-art VRP methods and identify strengths, weaknesses, and success factors.
• Highlight emerging research topics and future directions in unified heuristics, automatic heuristic design, and machine learning-assisted approaches.

Proposed method

• Classify VRP heuristics into three main categories: constructive, improvement, and metaheuristics, and review their methodologies and applications.
• Present a general algorithm framework for SOTA VRP methods and summarize their components and success factors.
• Discuss recent advances and practical applications of constructive, improvement, and metaheuristic methods, including algorithmic relationships and performance considerations.

Experimental results

Research questions

• RQ1 What are the main categories and methodologies of VRP heuristics and how have they evolved recently?
• RQ2 What is the SOTA framework that unifies different VRP heuristics across constructive, improvement, and metaheuristic classes?
• RQ3 What are the emerging topics (unified heuristics, automatic design, ML-assisted approaches) guiding future VRP research?

Key findings

• Constructive heuristics quickly produce feasible solutions but often with a gap from optimality across methods like nearest neighbor, insert, saving, and sweep.
• Improvement heuristics explore neighborhoods (intra-route and inter-route) to efficiently reach local optima, with trade-offs between computation and solution quality.
• Metaheuristics (single-solution and population-based) dominate VRP research, with ALNS and large neighborhood search showing strong practical impact; machine learning and RL are used to guide or select moves.
• There is increasing integration of machine learning techniques to guide improvements and search strategies, including learning cost reductions and dynamic selection of operators.
• The paper identifies three emerging topics: unified heuristics, automatic heuristic design, and machine learning-assisted heuristics, and discusses future directions.