Hallucination-Aware Generative Pretrained Transformer for Cooperative Aerial Mobility Control

2025-04-15

AI Summary

To address delivery failures in UAV last-mile logistics caused by battery depletion and path hallucinations (e.g., redundant node revisits), this paper proposes SafeGPT, a two-tiered framework. At the global level, a GPT-based planner allocates regional tasks; at the edge, an on-device GPT performs real-time path planning, augmented by a reinforcement learning–based safety filter—trained via PPO or SAC—to dynamically suppress unsafe actions. We introduce the first hallucination-aware mechanism, integrating semantic reasoning with formal safety constraints and leveraging a dual replay buffer for efficient experience reuse. Experimental results demonstrate that, compared to a pure-GPT baseline, SafeGPT improves delivery success rate by 18.7%, reduces battery consumption by 32.4%, and shortens average flight distance by 26.1%, significantly enhancing both operational reliability and energy efficiency.

Technology Category

• Planning, Routing, and Scheduling: Replanning and Plan Repair
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Natural Language Processing: Safety and Robustness

Application Category

• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI
• Search and Retrieval-Augmented AI: Agentic search
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

This paper proposes SafeGPT, a two-tiered framework that integrates generative pretrained transformers (GPTs) with reinforcement learning (RL) for efficient and reliable unmanned aerial vehicle (UAV) last-mile deliveries. In the proposed design, a Global GPT module assigns high-level tasks such as sector allocation, while an On-Device GPT manages real-time local route planning. An RL-based safety filter monitors each GPT decision and overrides unsafe actions that could lead to battery depletion or duplicate visits, effectively mitigating hallucinations. Furthermore, a dual replay buffer mechanism helps both the GPT modules and the RL agent refine their strategies over time. Simulation results demonstrate that SafeGPT achieves higher delivery success rates compared to a GPT-only baseline, while substantially reducing battery consumption and travel distance. These findings validate the efficacy of combining GPT-based semantic reasoning with formal safety guarantees, contributing a viable solution for robust and energy-efficient UAV logistics.

Problem

• Mitigate hallucinations in UAV control using GPT and RL
• Ensure safe UAV deliveries by overriding unsafe actions
• Improve energy efficiency and success rate in UAV logistics

Innovation

• Two-tiered GPT framework for UAV control
• RL safety filter prevents unsafe actions
• Dual replay buffer refines strategies