Team AIRLab POLIMI
Team members
Mirko Usuelli, Leonardo Gargani, Mattia Caon, Ettore Castelli, Simone Edmondo Bedini, Riccardo Serraino, Giuliano Livi, Giulio Cestele
Team captain’s name
Mirko Usuelli
Instructor(s)
Mirko Usuelli, Leonardo Gargani – Politecnico di Milano, Italy
Academic supervisors: Prof. Matteo Matteucci, Dr. Simone Mentasti
Participating in the FRE since
2022
Description of the team and robot
The AIRLab Team is part of one of the oldest research laboratories in AI and robotics in Italy at Politecnico di Milano. Through the Robotics course, the lab offers students the opportunity to apply their knowledge to real‒world agricultural challenges by participating in the Field Robot Event. The team, composed of MSc Computer and Automation Engineering students, develops autonomous robots using AI, computer vision, and perception techniques. Founded in 2022, it has quickly grown, earning several medals in recent years while helping train the next generation of robotics engineers.
Robot specifications
W x L x H (cm):
54.0 x 40.9 x 26.3
Weight (kg):
15
Commercial or prototype:
Prototype
Total no. of wheels / no. driven wheels:
4-wheel drive
Drivetrain concept/ max. speed (m/s):
3 m/s
Turning radius (cm):
0
Battery type / capacity (Ah):
12 Ah
Total motor power (W):
150 W
No. of sensors internal/
external:
Sensor type:
2x single-plane laser-scan (front/back)
2x rgb-d cameras (right/left)
1x frontal rgb stereo camera
Controller system software description:
We rely on a modular, dockerized architecture composed of interconnected submodules running ROS 2 Humble on Ubuntu 22.04. Our software stack integrates both C++ and Python: C++ is primarily used for core robotics functionalities such as navigation and low-level perception, while Python is adopted for the visual perception framework. The overall system is coordinated through finite state machines and behavior trees, selected according to the specific task requirements.
Controller system hardware description:
The robot is equipped with four independent wheels, enabling dynamic adaptation of its kinematics at runtime. Depending on the task, the system can switch between skid-steering, Ackermann, and omnidirectional configurations. The platform includes two computing units: a primary Intel Core i7-based system responsible for robot control, and a secondary NVIDIA Jetson unit dedicated to vision processing.
Short strategy description for navigation and applications:
Our navigation strategy is primarily based on LiDAR (laser-scan) data, using a custom approach that clusters crop vegetation along the sides of the robot. This enables robust row-following through goal-point tracking combined with geometry-based regularization, particularly effective in sparse scenarios. For turning maneuvers, the system leverages the geometric structure at the end of each row to estimate the robot’s orientation and executes a sequence of closed-loop motions. This ensures smooth curvature alignment with the crop layout before entering the next row.
These are the commercial team sponsors & partners:
NovaLabs
, Oversonic Robotics, Agricola Moderna