Connected Robotics Over 5G for Autonomous Networks
Coach Name
Juan Juan
EU Organization
EURECOM (France)
Members
- Omid Esrafilian
- David Gesbert
- Mohsen Ahadi
US Organization
Northeastern University (USA)
Members
- Tommaso Melodia
- Florian Kaltenberger
Project Overview
COBOTPLANET develops a foundational framework for intelligent, 5G-connected multi-robot networks, enabling autonomous drones to provide services such as connectivity extension, sensing, and user localization.
The project responds to a growing technological shift: wireless networks are no longer just infrastructure — they are evolving into adaptive robotic systems able to sense, learn, and collaborate in real time. However, existing technologies for robotics, wireless communication, and AI still operate in silos, lacking the interoperability needed for scalable and cooperative robotic networks.
COBOTPLANET bridges this gap by building the first cohesive architecture integrating 5G communication, robotic control, distributed intelligence, and real-world testbeds. It delivers new hardware for drone-mounted 5G nodes, a high-fidelity ray-tracing simulator, distributed multi-UAV coordination, and proof-of-concept demonstrations enabling decentralized robotic decision-making.
The results have broad societal potential, including emergency response, public safety, coverage restoration, and environmental monitoring.
Methods and approaches
Integrated 5G–Robotics Architecture with Real Hardware & Simulation
The team designed:
- a 5G base-station box for drones acting as aerial BSs,
- a 5G user-terminal box mountable on drones,
- a drone-enabled radio simulator with full ray-tracing capabilities, and
- a decentralized connected robotics platform using ROS 2 and cloud-assisted coordination.
This creates a unified environment linking communication, mobility, and sensing.
AI-driven User Localization Using 5G Signals
COBOTPLANET developed and tested a self-supervised 5G localization framework achieving <6m accuracy, even under non-line-of-sight conditions. It integrates:
- Channel Charting (CC),
- Time Difference of Arrival (TDoA) features,
- noisy-measurement masking,
- drone displacement priors.
The method outperformed state-of-the-art CC approaches in real-world 5G testbeds.
Key Achievements
Drone-mounted 5G base-station prototype (KPI1).
Drone-mounted 5G terminal prototype (KPI2).
Open-source ray-tracing UAV channel simulator in Python (KPI3).
Full technical documentation for simulator (KPI4).
Achieved 100 Mbps throughput beyond 20m (KPI5).
Demonstrated 50m+ 5G connectivity range (KPI6).
Decentralized connected robotics proof-of-concept using cloud migration (KPI7).
Drone localization with <6m accuracy, validated in real tests (KPI8).
Scientific paper submitted, detailing localization innovations (KPI9).
Robust testbed results using OpenAirInterface and ROS 2.
Recruitment of a research engineer who later became a PhD candidate.
Impact & Results
Scientific Impact
COBOTPLANET introduces important advancements in:
- radio-based localization,
- self-supervised learning for 5G positioning,
- integrated 5G–robotics experimentation,
- decentralized control for multi-UAV systems.
These contributions are formalized in a submitted journal paper.
Societal Impact
The technology can dramatically improve:
- disaster response (deployable connectivity, search & rescue),
- emergency communications,
- public safety,
- environmental sensing.
Economic & Industrial Impact
COBOTPLANET provides groundwork for next-generation autonomous 5G robotic systems, opening pathways for:
- telecom operators,
- UAV manufacturers,
- industrial automation,
- defense and security applications.
A startup is under consideration, supported by EURECOM’s TechForward incubator.
EU–US Collaboration
Partnership with Northeastern University will continue via:
- joint publications,
- new project proposals,
- workshops and seminars.
Publications and Open-Source Contributions
- UAV ray-tracing simulator (GitLab).
- Multiple technical reports hosted on Google Drive (hardware, performance, coverage).
- 5G localization paper (arXiv).
- Video evidence of drone localization accuracy.
- Documentation and results linked throughout report KPIs.
Future directions