Autonomous Multi-UAV Exploration and Rescue in Extreme Environments
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Autonomous navigation in extreme hazardous environments where human access is impossible remains a critical challenge for unmanned aerial vehicles (UAVs).This paper presents a novel Hierarchical Multi-Agent Deep Reinforcement Learning (HMA-DRL) framework for coordinated UAV swarms capable of exploring and navigating through disaster zones, confined underground spaces, and chemically contaminated areas.Building upon recent advances in Deep Proximal Policy Optimization (D-PPO) and multi-agent coordination, our approach integrates Enhanced Perception Networks (EPN), Adaptive Communication Protocols (ACP), and Risk-Aware Decision-Making (RADM) modules.The system combines monocular depth estimation, thermal imaging, and gas detection sensors for comprehensive environmental awareness.Experimental validation in simulated disaster scenarios demonstrates 89% improvement in exploration coverage, 76% reduction in collision rates, and successful navigation through spaces with visibility less than 2 meters.Realworld testing in abandoned underground facilities shows the system's capability to autonomously map and navigate 450+ meter tunnel networks while maintaining swarm cohesion and avoiding hazardous gas concentrations above 50 ppm.
Fields of Study & Subjects
Reinforcement Learning in Robotics
UAV Applications and Optimization
Modular Robots and Swarm Intelligence
Publication details
- Journal / source
- Advances in intelligent systems research/Advances in Intelligent Systems Research
- Pages
- 600–611
- Publication date
- 1 April 2026
- Language
- EN
- Document type
- Book chapter