Challenge 1
GenAI-Based Natural Language Mission Generator for Autonomous Robots in Agriculture _ Intellimech
Team:
Motiq (Motiq Labs), Croatia
Proposal title:
Natural-Language Mission Planning for Vineyard Robots
Organisation
Motiq is a Croatian team of engineers with a shared interest in robotics and AI. Marijan Miletić, Petar Deverić and Ivan Deverić bring together experience in mechatronics, software infrastructure and connecting AI models to robots. Two team members took up teaching posts at Zagreb University of Applied Sciences (TVZ) in October 2026. The team is exploring how robotics and AI can be used in agriculture. They entered AI-BOOST as an independent team and founded Motiq Labs after being selected in the Spark phase. Website: https://www.motiq.hr/
Project info
Motiq’s proposed system would let a vineyard operator give a robot spoken instructions. Before the robot acts, the system would check whether the proposed mission is within the robot’s capabilities. The team’s starting point was a simple test in which an instruction to remain still nevertheless produced joint movements.
At the heart of the proposal is a catalogue that describes what the robot can do. If an instruction calls for an action beyond those capabilities, the system would decline the task and explain why, so the operator can revise the instruction.
A vineyard operator might ask the robot to “inspect the fifth row”. To plan the inspection, the system would need to identify the right row and work out how the robot could carry out the task. During the Advance phase, the team will evaluate how reliably the system translates instructions like this into robot missions.
Details
Letting growers give instructions in everyday language could make vineyard robots easier to use. The challenge is that a plan can sound plausible even when it includes an action the robot cannot carry out. Motiq’s proposal addresses this by checking each mission against the robot’s capabilities before execution.
Motiq proposes using a capability catalogue to describe what the robot can do. If the system finds a mission that fits those capabilities, it would pass the checked plan to the robot. If it cannot find one, it would explain why, so the operator can revise the instruction before the robot acts.
Following its selection in the Spark phase of AI-BOOST Challenge 1, Motiq is developing a proof of concept in the Advance phase with Consorzio Intellimech and JOiiNT LAB. The team will assess whether the system understands different ways of phrasing the same request, produces a workable mission within a practical response time, and rejects tasks the robot cannot perform.
