Algorithmic Prompt-Augmentation for Efficient LLM-Based Heuristic Design for A* Search

TL;DR

This paper presents A-CEoH, a novel prompt augmentation method leveraging LLMs and evolutionary algorithms to automate heuristic design for A* search, demonstrating superior performance in logistics and puzzle domains.

Contribution

It introduces A-CEoH, a domain-agnostic prompt augmentation strategy that enhances heuristic generation for A* search using LLMs and evolutionary frameworks.

Findings

A-CEoH significantly improves heuristic quality.

Outperforms expert-designed heuristics.

Effective in logistics and puzzle domains.

Abstract

Heuristic functions are essential to the performance of tree search algorithms such as A*, where their accuracy and efficiency directly impact search outcomes. Traditionally, such heuristics are handcrafted, requiring significant expertise. Recent advances in large language models (LLMs) and evolutionary frameworks have opened the door to automating heuristic design. In this paper, we extend the Evolution of Heuristics (EoH) framework to investigate the automated generation of guiding heuristics for A* search. We introduce a novel domain-agnostic prompt augmentation strategy that includes the A* code into the prompt to leverage in-context learning, named Algorithmic - Contextual EoH (A-CEoH). To evaluate the effectiveness of A-CeoH, we study two problem domains: the Unit-Load Pre-Marshalling Problem (UPMP), a niche problem from warehouse logistics, and the classical sliding puzzle problem (SPP). Our computational experiments show that A-CEoH can significantly improve the quality of the generated heuristics and even outperform expert-designed heuristics.