Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A $1000\times$ Faster LLM-enhanced Algorithm For Path Planning in Large-scale Grid Maps

Junlin Zeng, Xin Zhang, Xiang Zhao, Yan Pan

TL;DR

This work tackles path planning on large-scale grid maps by enhancing LLM-guided planning with iLLM-A*. It introduces three mechanisms: hash-based CLOSED lists, selective OPEN-list heuristic updates, and a two-stage collision check to accelerate A*; incremental, learning-guided LLM waypoint generation; and empirical waypoint selection to prune redundancy. Across diverse map sizes and novel obstacle scenarios, iLLM-A* delivers over 1000x speedups (up to 2349.5x), up to 58.6% memory savings, and, on average, shorter and more stable paths closer to the optimal. The approach significantly improves scalability and practicality of LLM-enhanced path planning for large grids, enabling real-time planning in complex environments.

Abstract

Path planning in grid maps, arising from various applications, has garnered significant attention. Existing methods, such as A*, Dijkstra, and their variants, work well for small-scale maps but fail to address large-scale ones due to high search time and memory consumption. Recently, Large Language Models (LLMs) have shown remarkable performance in path planning but still suffer from spatial illusion and poor planning performance. Among all the works, LLM-A* \cite{meng2024llm} leverages LLM to generate a series of waypoints and then uses A* to plan the paths between the neighboring waypoints. In this way, the complete path is constructed. However, LLM-A* still suffers from high computational time for large-scale maps. To fill this gap, we conducted a deep investigation into LLM-A* and found its bottleneck, resulting in limited performance. Accordingly, we design an innovative LLM-enhanced algorithm, abbr. as iLLM-A*. iLLM-A* includes 3 carefully designed mechanisms, including the optimization of A*, an incremental learning method for LLM to generate high-quality waypoints, and the selection of the appropriate waypoints for A* for path planning. Finally, a comprehensive evaluation on various grid maps shows that, compared with LLM-A*, iLLM-A* \textbf{1) achieves more than $1000\times$ speedup on average, and up to $2349.5\times$ speedup in the extreme case, 2) saves up to $58.6\%$ of the memory cost, 3) achieves both obviously shorter path length and lower path length standard deviation.}

A $1000\times$ Faster LLM-enhanced Algorithm For Path Planning in Large-scale Grid Maps

TL;DR

This work tackles path planning on large-scale grid maps by enhancing LLM-guided planning with iLLM-A*. It introduces three mechanisms: hash-based CLOSED lists, selective OPEN-list heuristic updates, and a two-stage collision check to accelerate A*; incremental, learning-guided LLM waypoint generation; and empirical waypoint selection to prune redundancy. Across diverse map sizes and novel obstacle scenarios, iLLM-A* delivers over 1000x speedups (up to 2349.5x), up to 58.6% memory savings, and, on average, shorter and more stable paths closer to the optimal. The approach significantly improves scalability and practicality of LLM-enhanced path planning for large grids, enabling real-time planning in complex environments.

Abstract

Path planning in grid maps, arising from various applications, has garnered significant attention. Existing methods, such as A*, Dijkstra, and their variants, work well for small-scale maps but fail to address large-scale ones due to high search time and memory consumption. Recently, Large Language Models (LLMs) have shown remarkable performance in path planning but still suffer from spatial illusion and poor planning performance. Among all the works, LLM-A* \cite{meng2024llm} leverages LLM to generate a series of waypoints and then uses A* to plan the paths between the neighboring waypoints. In this way, the complete path is constructed. However, LLM-A* still suffers from high computational time for large-scale maps. To fill this gap, we conducted a deep investigation into LLM-A* and found its bottleneck, resulting in limited performance. Accordingly, we design an innovative LLM-enhanced algorithm, abbr. as iLLM-A*. iLLM-A* includes 3 carefully designed mechanisms, including the optimization of A*, an incremental learning method for LLM to generate high-quality waypoints, and the selection of the appropriate waypoints for A* for path planning. Finally, a comprehensive evaluation on various grid maps shows that, compared with LLM-A*, iLLM-A* \textbf{1) achieves more than speedup on average, and up to speedup in the extreme case, 2) saves up to of the memory cost, 3) achieves both obviously shorter path length and lower path length standard deviation.}

Paper Structure

This paper contains 30 sections, 4 equations, 4 figures, 9 tables.

Table of Contents

  1. Introduction
  2. Limitation Analysis for LLM-A*
  3. Limitation 1: Inefficient Implementation of A*.
  4. Limitation 2: Global High Memory Cost.
  5. Limitation 3: Inherent Limitation of LLM.
  6. Design of iLLM-A*
  7. Optimization of A*
  8. Optimization of The CLOSED List.
  9. Optimization of The OPEN List.
  10. Collision Detection Acceleration.
  11. Waypoint Generating by Incremental Learning-Based LLM
  12. Prompt Template.
  13. Incremental Learning Based Few-shot Example Augmentation.
  14. Task Instructions.
  15. Appropriate Waypoint Selection
  16. ...and 15 more sections

Figures (4)

  • Figure 1: Search Time and Memory of LLM-A* and A* Given Different Map Size.
  • Figure 2: Two-Stage Collision Detection.
  • Figure 3: CDF of Path Length For LLM-A* and iLLM-A*.
  • Figure 4: Basic Concept of the Two New Kinds of Obstacles.