Context-Aware Autonomous Drone Navigation Using Large Language Models (LLMs)

Abstract

In this paper, a novel large language model (LLM)-based context-aware autonomous drone navigation algorithm is presented. This approach demonstrates the capability of LLMs to navigate complex environments by balancing multisensor objectives with a weighted prioritization system. Specifically, we incorporate weights for the goals of obstacle avoidance, weather adaptation, and mission completion. The model's performance is tested under six progressively intricate scenarios in extensive simulations focused on path efficiency, completion time, and success rate. Results indicate that the LLM-based context-aware navigation algorithm achieves 94% success rate in simple environment in a moderate weather conditions with reasonable efficiency, and surpasses expectations in the advanced AI driven obstacle reasoning. These results illustrate the emerging strengths of LLMs for autonomous navigation and its potential utilization in situation where environmental conditions change dynamically.