Agent in the Sky: Intelligent Multi-Agent Framework for Autonomous HAPS Coordination and Real-World Event Adaptation

Dezheng Han, Anbang Zhang, Ruxiao Chen, Chenyuan Feng, Shuaishuai Guo

Keywords: Multi-Agent, Large Language Model, HAPS Network Optimization, AI Agent

TL;DR: Using cooperative multiple LLM-based agents for environement and event-adaptive network optimization

Abstract: High Altitude Platform Station (HAPS) offers significant flexibility for dynamic adaptability and efficient user coverage. However, achieving high levels of automation in HAPS systems is fraught with challenges, particularly in comprehending complex environments and processing natural language inputs essential for autonomous operations. Existing methods, such as reinforcement learning, are task-specific and lack the ability to integrate broader environmental information. To address these limitations, we propose an Autonomous Coverage Multi-Agent (ACMA) framework, which leverages Large Language Models (LLMs) to enhance coverage through intelligent coordination of HAPS. By incorporating techniques like in-context learning, fine-tuning, and tool-calling, our framework enables agents to understand and respond to environmental cues and natural language instructions effectively. Simulation results demonstrate that the ACMA system outperforms traditional methods in coordinating coverage, adeptly managing dynamic incidents and maximizing user coverage. Compared to traditional approaches, ACMA exhibits higher intelligence and autonomy, paving the way for more adaptable and efficient HAPS systems in real-world scenarios.

Submission Number: 7