HMACE: Heterogeneous Multi-Agent Collaborative Evolution for Combinatorial Optimization

TL;DR

HMACE introduces a multi-agent collaborative framework for heuristic search in combinatorial optimization, improving solution quality and efficiency over existing methods.

Contribution

This work presents HMACE, a heterogeneous multi-agent system that decomposes heuristic search into specialized roles, enabling more effective exploration and memory-guided optimization.

Findings

HMACE outperforms state-of-the-art baselines on TSP and BPP with lower average gaps.

HMACE requires significantly fewer tokens than comparable methods.

Extensive evaluations show HMACE achieves a favorable quality-efficiency trade-off.

Abstract

Large Language Models have recently emerged as a promising paradigm for automated heuristic design for NP-hard combinatorial optimization problems. Despite this progress, existing LLM-based methods typically rely on monolithic workflows constrained by rigid templates, thereby restricting memory-guided exploration and triggering premature convergence to local optima. To design an autonomous and collaborative architecture, we introduce HMACE, a Heterogeneous Multi-Agent Collaborative Evolution framework that reconceptualizes heuristic search as an organizational design problem. HMACE decomposes each evolutionary generation into an autonomous, role-specialized loop with four coordinated agents: a Proposer for strategy exploration, a Generator for executable heuristic synthesis, an Evaluator for empirical assessment, and a Reflector for archive-backed memory update. By coupling behavior-aware retrieval, lightweight candidate filtering, and fitness-grounded archive updates, HMACE guides the search toward diverse and promising heuristic behaviors while avoiding redundant evaluations. Extensive evaluations on representative COPs, including TSP, Online BPP, MKP, and PFSP, show that HMACE achieves a favorable quality-efficiency trade-off compared to state-of-the-art single-agent and multi-agent baselines. In the matched LLM-driven reference comparison, HMACE achieves the lowest average gaps on TSP and Online BPP (0.464\% and 0.223\%, respectively), while requiring only 0.13M and 0.42M tokens for the two tasks, substantially fewer than the compared baselines.