The project aims to convert human demonstrations and verbal instructions into machine-readable task sequences that can be executed by coordinated teams of robots operating in real-world environments.

This initial effort focuses on agricultural work and takes a unique human-robot approach that might be applied to manufacturing in the future:

According to a release, "farmers will show and explain their routines [to the robots]. Short demonstrations and spoken narrative notes will be turned into step-by-step instructions that robots can follow. A small on-site computer will coordinate teams of lightweight ground vehicles and drones that scout fields, flag issues and carry out tasks under strict safety limits and with human oversight. A simple app will allow farmers to review and adjust the robots’ work."

The effort reflects a broader shift toward physical AI, where robots learn from human actions and adapt to changing conditions rather than relying solely on predefined code and structured automation.

Forth UK, a British engineering company known for developing robotic systems for nuclear decommissioning and other hazardous industrial environments, will serve as an industry partner and manufacture project prototypes.

The company said many of the technologies originally developed for remote operation, inspection and automation in safety-critical industrial environments are increasingly being adapted for broader commercial robotics applications.

The FARMAR project has received €1.79 million in funding through the Horizon Europe research program and is being coordinated by Durham University’s Department of Computer Science.