Pilot Case 3
Robotic fertilisation management for leafy vegetables in open field conditions in Italy
Location: Experimental and commercial farms in Italy
Overview
In Pilot Case 3, AgRibot is developing a smarter way to fertilise lettuce in open fields. Instead of treating the entire field the same, this pilot case focuses on understanding how each part of the crop is actually growing and fertilising accordingly.
To achieve this, a robotic platform moves through the field and scans the plants using RGB-D cameras and other sensors. These sensors capture information about plant size, structure, and growth. AI models then analyse this data to estimate plant biomass and overall condition.
This allows the system to identify where plants may need more nutrients and where they are already performing well. The goal is to avoid unnecessary applications, use fertiliser more efficiently, and support more uniform crop development.
The pilot is carried out in Bari, Italy, by CNR, Politecnico di Bari, and Università degli Studi di Bari, under real open-field conditions.
POLIBOT
The robotic platform, named POLIBOT, is a robotic farmer developed by Politecnico di Bari. It is designed specifically for open-field operations and can move steadily across uneven terrain thanks to its tracked system, which also helps reduce soil compaction.
POLIBOT carries a combination of 2D and 3D sensors that collect detailed crop data while driving through the field. It can survey around one hectare in roughly 40 minutes, operating at speeds of 1–2 metres per second. Its structure is built to remain stable during movement, ensuring reliable measurements.
Rather than replacing the farmer, POLIBOT acts as a field assistant, collecting detailed information that supports better fertilisation decisions.
Timeline
Field activities started in spring 2025 with the first data collection campaign in Bari. During April and May, the first full dataset was gathered from lettuce grown under open-field conditions.
A second campaign followed between October and December 2025, this time using POLIBOT directly in the field. The additional dataset helped improve the reliability of the AI models under different seasonal conditions.
Model development is ongoing, with continuous refinement to ensure stable and accurate plant trait estimation across growth stages.
Validation
The system is tested under real farming conditions to ensure it performs beyond controlled environments. Plant traits estimated by the AI models are compared with traditional crop assessments to verify accuracy.
At the same time, the pilot evaluates practical impact: more precise fertiliser use, fewer unnecessary applications, improved crop uniformity, and reliable robot performance in real fields.
AR Integration
Augmented reality tools are used to make the collected crop data easier to understand. Fertilisation recommendations and plant condition indicators can be visualised through AR-supported field models.
By combining robotics, AI, and AR visualisation, Pilot Case 3 aims to translate detailed field measurements into practical insights that farmers can confidently act upon.