Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

New Adaptive Mechanism for Large Neighborhood Search using Dual Actor-Critic

Shaohua Yu, Wenhao Mao, Zigao Wu, Jakob Puchinger

TL;DR

The paper tackles the limited effectiveness of traditional ALNS adaptive mechanisms by introducing a Dual Actor-Critic (DAC) framework that jointly optimizes destroy and repair operator selection. It leverages Graph Neural Networks to construct state representations for both phases and frames destroy/repair as two coupled Markov Decision Processes evaluated by a shared critic. The approach yields improved solution quality, particularly on VRPTW, and demonstrates strong transferability through a zero-shot DAC-T variant that trains on small instances and performs well on larger or unseen ones. This work advances practical logistics optimization by enabling faster, more scalable, and deployment-friendly adaptive search in complex routing problems.

Abstract

Adaptive Large Neighborhood Search (ALNS) is a widely used heuristic method for solving combinatorial optimization problems. ALNS explores the solution space by iteratively using destroy and repair operators with probabilities, which are adjusted by an adaptive mechanism to find optimal solutions. However, the classic ALNS adaptive mechanism does not consider the interaction between destroy and repair operators when selecting them. To overcome this limitation, this study proposes a novel adaptive mechanism. This mechanism enhances the adaptability of the algorithm through a Dual Actor-Critic (DAC) model, which fully considers the fact that the quality of new solutions is jointly determined by the destroy and repair operators. It effectively utilizes the interaction between these operators during the weight adjustment process, greatly improving the adaptability of the ALNS algorithm. In this mechanism, the destroy and repair processes are modeled as independent Markov Decision Processes to guide the selection of operators more accurately. Furthermore, we use Graph Neural Networks to extract key features from problem instances and perform effective aggregation and normalization to enhance the algorithm's transferability to different sizes and characteristics of problems. Through a series of experiments, we demonstrate that the proposed DAC-ALNS algorithm significantly improves solution efficiency and exhibits excellent transferability.

New Adaptive Mechanism for Large Neighborhood Search using Dual Actor-Critic

TL;DR

The paper tackles the limited effectiveness of traditional ALNS adaptive mechanisms by introducing a Dual Actor-Critic (DAC) framework that jointly optimizes destroy and repair operator selection. It leverages Graph Neural Networks to construct state representations for both phases and frames destroy/repair as two coupled Markov Decision Processes evaluated by a shared critic. The approach yields improved solution quality, particularly on VRPTW, and demonstrates strong transferability through a zero-shot DAC-T variant that trains on small instances and performs well on larger or unseen ones. This work advances practical logistics optimization by enabling faster, more scalable, and deployment-friendly adaptive search in complex routing problems.

Abstract

Adaptive Large Neighborhood Search (ALNS) is a widely used heuristic method for solving combinatorial optimization problems. ALNS explores the solution space by iteratively using destroy and repair operators with probabilities, which are adjusted by an adaptive mechanism to find optimal solutions. However, the classic ALNS adaptive mechanism does not consider the interaction between destroy and repair operators when selecting them. To overcome this limitation, this study proposes a novel adaptive mechanism. This mechanism enhances the adaptability of the algorithm through a Dual Actor-Critic (DAC) model, which fully considers the fact that the quality of new solutions is jointly determined by the destroy and repair operators. It effectively utilizes the interaction between these operators during the weight adjustment process, greatly improving the adaptability of the ALNS algorithm. In this mechanism, the destroy and repair processes are modeled as independent Markov Decision Processes to guide the selection of operators more accurately. Furthermore, we use Graph Neural Networks to extract key features from problem instances and perform effective aggregation and normalization to enhance the algorithm's transferability to different sizes and characteristics of problems. Through a series of experiments, we demonstrate that the proposed DAC-ALNS algorithm significantly improves solution efficiency and exhibits excellent transferability.
Paper Structure (28 sections, 4 equations, 3 figures, 8 tables, 2 algorithms)

This paper contains 28 sections, 4 equations, 3 figures, 8 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Adaptive Large Neighborhood Search
  4. Deep reinforcement learning in ALNS
  5. Methodology
  6. DAC-ALNS algorithm architecture
  7. Markov Decision Processes
  8. State Space
  9. State Representation (CVRP)
  10. State Representation (VRPTW)
  11. Action Space
  12. Reward Function
  13. Numerical Experiments
  14. Experimental Setup
  15. Datasets and Benchmarks
  16. ...and 13 more sections

Figures (3)

  • Figure 1: Actor-Critic algorithm diagram
  • Figure 2: Operator weight adjustment mechanism based on DAC
  • Figure 3: Schematic diagram of the operational logic for ALNS, AC-ALNS, and DAC-ALNS.