Research Output

Publications

Peer-reviewed papers from the NAMUR team on human–multirobot interaction, LLM-based coordination, and search & rescue robotics.

One Operator, Many Robots: Natural-Language Command of UAV–UGV Teams for Search, Victim Detection, and Rescue Under review

Alejandro Jarabo-Peñas, Juan Bravo-Arrabal, Javier Serón-Barba, D. Lin-Yang, F. Pastor, R. Ladig, J. J. Fernández-Lozano, Alfonso García-Cerezo, Anders Lyhne Christensen

IEEE International Symposium on Safety, Security, and Rescue Robotics (SSRR) 2026

A real-world SAR extension of Swarm-Steward that connects multi-UAV search, live aerial victim detection, geolocation, operator confirmation, and UGV deployment — one operator voice-commands the whole team. Evaluated in 12 outdoor experiments over a 47,362 m² unstructured area with one-, two-, and four-UAV teams, achieving best victim-localization errors of 1.21–2.29 m, with feedback from a professional firefighter.

Swarm-Steward: Scalable and Reliable Natural-Language Coordination of Autonomous Aerial and Ground Robots

Alejandro Jarabo-Peñas, Juan Bravo-Arrabal, Edouard G. A. Rolland, Anders Lyhne Christensen

2026 International Conference on Unmanned Aircraft Systems (ICUAS) 2026

A platform-agnostic hierarchical LLM multi-agent system that lets a non-expert operator coordinate swarms of aerial and ground robots through natural language. It separates planning and grounding from a safety-gated deterministic execution stage, uses dual retrieval-augmented generation (map-feature + telemetry), keeps token cost near-constant from 5 to 500 drones, achieves 92.9% task success, and validates sim-to-real transfer on a DJI Mini 4 Pro swarm.

Real-World Deployment of an LLM-Enabled Voice-Commanded UGV for Logistics in SAR Missions

Alejandro Jarabo-Peñas, Juan Bravo-Arrabal, D. Lin-Yang, F. Pastor, R. Ladig, J. J. Fernández-Lozano, Anders Lyhne Christensen, A. García-Cerezo

2025 IEEE International Symposium on Safety, Security, and Rescue Robotics (SSRR) 2025

Deploys an LLM-enabled voice-command interface on a UGV for logistics support in search-and-rescue missions, letting operators without robotics expertise issue natural-language voice commands that the robot grounds in real-time mission data for safe, context-aware navigation.

SAREnv: An Open-Source Dataset and Benchmark Tool for Informed Wilderness Search and Rescue Using UAVs

Kasper Andreas Rømer Grøntved, Alejandro Jarabo-Peñas, Sid Reid, Edouard George Alain Rolland, Matthew Watson, Arthur Richards, Steve Bullock, Anders Lyhne Christensen

Drones (MDPI) 2025

An open-access dataset and evaluation framework for benchmarking UAV-based wilderness search-and-rescue planning algorithms — shared infrastructure developed under the NAMUR project umbrella (in collaboration with the WildDrone/HERD teams).