LAPF: LLM-Agent-Based Path Finder Using the UAVScenes Dataset

2026-08-15

AI Summary

This study addresses the limited generalization of outdoor UAV navigation and functional constraints of LLM-based agents by proposing the LAPF framework. This method establishes a closed-loop cognitive architecture integrating perception, memory, and planning, which couples hazard detection with bounded corrective actions via chain-of-thought reasoning and dynamically optimizes waypoint decisions based on environmental feedback. Evaluated on the UAVScenes dataset, the framework achieves 98.1% path efficiency, representing a 15.6% improvement over standard CoT prompting, while ensuring zero collisions and stable near-target flight. These results demonstrate that LAPF significantly enhances adaptive navigation capabilities in complex urban-scale environments, effectively overcoming current limitations in autonomous aerial systems.

Abstract

Uncrewed aerial vehicles (UAVs) are increasingly deployed for autonomous navigation in complex outdoor environments, where dynamic conditions and mission requirements require intelligent adaptive decision-making. Existing optimization-based, Machine Learning (ML), and Reinforcement Learning (RL) approaches often rely on predefined models or task-specific training, limiting their generalization and adaptability in uncertain scenarios. Recent Large Language Model (LLM)-assisted approaches offer promising reasoning capabilities but remain constrained by limited agentic functionality, including insufficient memory, planning, and tool interaction mechanisms.This paper proposes an LLM-Agent-Based Path Finder (LAPF) framework for autonomous UAV navigation in town-scale outdoor environments. LAPF extends LLM-assisted navigation by integrating perception, memory, planning, and action modules into a closed-loop cognitive architecture. The proposed agent leverages prior navigation experiences, performs Chain-of-Thought (CoT) reasoning, couples each detected hazard to a bounded corrective action, and dynamically refines waypoint decisions based on environmental feedback.The three independent trials per method demonstrate that LAPF achieves mean path lengths of 512.83 m and 506.37 m, compared to the straight-line optimum of 497.33 m, corresponding to path length reductions of 17.2% and 15.6% relative to CoT prompting and absolute path efficiencies of 97.1% and 98.1% in open-field and obstacle-injected scenarios, respectively. Furthermore, LAPF is the only evaluated approach that couples every detected hazard to a bounded, metric-neutral corrective action while maintaining near-goal stability, with zero clamp events in both scenarios, whereas CoT prompting increases from 9.7 to 14.0 events.

Problem

• UAV autonomous navigation
• Large Language Models
• LLM-Agent
• outdoor environments

Innovation

• LLM-Agent
• Closed-loop Cognitive Architecture
• Chain-of-Thought Reasoning
• Bounded Corrective Action
• Autonomous UAV Navigation