Cross domain Persistent Monitoring for Hybrid Aerial Underwater Vehicles
A single Deep Reinforcement Learning policy controls hybrid drones using both Lidar and Sonar data.
A research team from Brazilian institutions including Ricardo B. Grando, Victor A. Kich, and colleagues has published a groundbreaking paper on "Cross domain Persistent Monitoring for Hybrid Aerial Underwater Vehicles." The work addresses a significant robotics challenge: creating autonomous systems that can seamlessly operate in both aerial and underwater environments. Hybrid Unmanned Aerial Underwater Vehicles (HUAUVs) represent a new class of platforms capable of applications like infrastructure inspection, environmental mapping, and search and rescue in coastal or maritime disaster scenarios. However, developing control methodologies has been difficult due to the radically different dynamics and sensor requirements between air (using Lidar) and water (using Sonar).
The team's technical breakthrough involves combining Deep Reinforcement Learning (DRL) with Transfer Learning to enable cross-domain adaptability. They developed a shared DRL architecture that trains on both Lidar sensor data for aerial navigation and Sonar data for underwater operation, demonstrating that a single, unified AI policy can effectively control the vehicle in both domains. The framework is designed to handle environmental uncertainty and track multiple mobile targets, laying the groundwork for scalable autonomous monitoring solutions. This research, accepted to the Brazilian Conference on Robotics 2026, represents a major step toward versatile robotic systems that can transition between mediums without needing completely separate control systems, potentially revolutionizing maritime surveillance, offshore infrastructure maintenance, and environmental monitoring.
- Uses a unified Deep Reinforcement Learning policy trained on both Lidar (air) and Sonar (water) data
- Enables Hybrid Unmanned Aerial Underwater Vehicles (HUAUVs) to perform persistent monitoring across domains
- Handles environmental uncertainty and tracks multiple mobile targets for applications like search & rescue
Why It Matters
Enables single autonomous systems for maritime surveillance, offshore inspections, and disaster response without domain switches.