


European agriculture is rapidly evolving — with increasing pressure for automation, sustainability, and labor efficiency. Inria plays a central role in enabling that transformation through advanced robotics R&D.
At its Sophia Antipolis campus, Inria leads the French node of the AgriFood TEF (Testing and Experimentation Facility) initiative, offering real-world infrastructure for robotics and AI experimentation in agriculture.
The problem :
Inria needed a stable, autonomous, and scalable network that would allow teams to test, manage, and operate AI-powered farming machines in real conditions — without depending on public carriers.
We’re not just testing this in theory. We’re putting it to work in real fields, with real robots, solving real problems.
Philippe Martinet, Research Director, Inria
To address these challenges, Inria deployed the Firecell Pegasus Private 5G system, creating a high-performance, field-grade wireless network tailored for robotics.
Key elements of the solution include:
The system enables:
Reliable communication over uneven terrain, dense vegetation, and GPS-obstructed zones.
Total network control: Inria manages provisioning, configuration, monitoring — without relying on telecom operators.
Supports bandwidth-heavy applications such as real-time HD video, LiDAR scans, and autonomous pathfinding.
Under French regulation, Inria licenses its own spectrum via ARCEP — a major enabler for research independence.
The Firecell deployment has transformed Inria’s robotic testbed into one of the most advanced AgriTech experimentation sites in Europe. Key outcomes include:
The platform supports everything from mobility algorithms to embedded firmware — enabling both fundamental research and applied innovation.
Inria plans to expand the testbed to support:
As a foundational node of the EU AgriFood TEF, Inria is not just supporting research — it’s building a scalable reference model for the farms of the future.
What we really need is a stable, resilient connection that works in the field. Private 5G helps us do that — and gives us a real architecture to scale.
Malek Aifa, Research Engineer, Inria