Open Positions - Master Thesis
Application Open: 08/12/2025
Application Deadline: 18/01/2025
Radar Data Extraction and Processing for Mobile Robots Odometry
Supervisor: Prof. Marcello Chiaberge, G. Audrito, M. Martini, M. Ambrosio
Skills: Mandatory: Basic Python Recommended: ROS 2, PyTorch
Code: 2026PIC01
Topic: Sensors / Perception
Description: The thesis activity will be carried out at PIC4SeR (PoliTO) and supervised by Leonardo Lab (Genova). This master’s thesis project focuses on the development and implementation of methods for radar data extraction and processing to enhance odometry estimation in mobile robots. The work involves designing algorithms for signal filtering and feature extraction, and integrating the processed data into an odometry pipeline. The objective is to improve the robot’s localization accuracy and robustness, particularly in environments where traditional sensors like cameras or LiDAR are limited. The project includes both software development and experimental validation.
VR Collaborative Manipulation Tasks for Robotic Harvesting in Agriculture
Supervisor: Prof. Marcello Chiaberge, B. Tuberga, U. Albertin, M. Ambrosio
Skills: Mandatory: Basic Python Recommended: ROS 2, MoveIt, C, Unity
Code: 2026PIC02
Topic: Agri-Robotics, VR, Manipulation
Description: Integrate a robotic control and learning framework in a Virtual Reality environment to let the robot learn from the interaction with humans. The Virtual Reality environment can be used to generate task demonstration data to teach the robot how to solve complex tasks (fruit harvesting, pruning…). The goal of the project is to identify and learn optimal control policy for the robot to perform agricultural tasks, and also to define guidelines for training human operators to cooperate with the robot. The simulated scene should be develop in Unity, good programming skills or background experience with simulators is recommended (C, Python). The target control framework adopted is MoveIt2.
Anti-slippage wheeled robot control with Physics-Informed Neural Network
Supervisor: Prof. Marcello Chiaberge, M. Martini, C. Cena, G. Franchini
Skills: Mandatory: Basic Python Recommended: ROS 2, PyTorch
Code: 2026PIC03
Topic: Wheeled Robots
Description: Wheels slippage is often affecting robot odometry and localization in both indoor and outdoor environments. Precisely estimating real platform velocity through wheel’s encoder data requires complex dynamic models that is often not possible to describe. The project aims to develop a data-driven solution to learn the correct dynamics of the robot and compute physics-aware commands for robot’s wheels to avoid slippage. The project include software development in PyTorch and experimental activity with real platform in the lab.
Multi-robot UWB Localization Benchmark
Supervisor: Prof. Marcello Chiaberge, G. Audrito, M. Martini, S. Primatesta
Skills: Mandatory: Basic Python Recommended: ROS 2, C++
Code: 2026PIC04
Topic: Multi-robot Localization
Description: The project aims to collect a wide and rich dataset for multi-robot UWB localization, experimenting trajectories in different environments (indoor & outdoor) with both UGVs and UAVs. A refined ground truth signal should be coupled with standard sensors data (UWB ranges) to perform a precise evaluation and comparison of state-of-the-art localization approaches.
Planning for Lunar Exploration with Autonomous Rover
Supervisor: Prof. Marcello Chiaberge, M. Di Paola, G. Audrito
Skills: Mandatory: Basic Python Recommended: ROS 2, C++
Code: 2026PIC05
Topic: Space Robotics
Description: The project aims to develop a planning algorithm for space robotics, with the goal of deploying critical payloads (transmitter-receiver anchors) on the Lunar surface to enhance the accurate localization of the rover in the area. An optimal trade-off should be found among multiple objectives (exploration, obstacle/risk avoidance, shape of anchors distribution, distance, area). The planning framework will be developed with a Factor Graph approach using the standard GTSAM library.
Visual autonomous Navigation for Horticulture
Supervisor: Prof. Marcello Chiaberge, A. Navone, U. Albertin, F. Aisa
Skills: Mandatory: Python Recommended: ROS 2, C++
Code: 2026PIC06
Topic: Agri-Robotics, CV
Description: The project aims to develop a robust algorithm that enables autonomous navigation on a low plant beds.
1) State of the art of Autonomous Navigation algorithms and Datasets available in literature
2) Develop a custom robust algorithm
3) Test the algorithm in a real field environment
4) Implement further improvements based on the advancements made by the student.
Plant Segmentation and Classification for Autonomous Weeding in Horticulture
Supervisor: Prof. Marcello Chiaberge, A. Navone, U. Albertin, F. Aisa
Skills: Mandatory: Python Recommended: ROS 2, C++
Code: 2026PIC07
Topic: Agri-Robotics, CV
Description: Automatic plant recognition using AI and Computer Vision techniques to perform localized operations. The idea is to collect or use public datasets to automatically recognize specific plants within a low plant bed.
1) State of the art of Computer Vision techniques and Datasets available in literature for plant recognition
2) Develop a custom robust algorithm able to recognize plants in different environments, from well-structured fields to unstructured ones.
3) Test the algorithm in a real field environment
4) Implement further improvements based on the advancements made by the student.
SLAM-based Mapping and Row Following for Autonomous Drone Navigation in Vineyards
Supervisor: Prof. Marcello Chiaberge, S. Primatesta, G. Audrito
Skills: Mandatory: Python Recommended: ROS 2, C++
Code: 2026PIC08
Topic: Agri-Robotics, CV
Description: This thesis aims to develop a local perception–based navigation system enabling autonomous drone operation within vineyard rows. By generating accurate environmental maps in real time, the system will support precise, automated spraying processes and enhance the efficiency and safety of vineyard management. The aim of this thesis:
1)Map the vineyard environment
2)Localize the UAV during flight
3)Detect and extract vineyard rows
4)Support autonomous row-following navigation
Low-cost LiDAR Odometry in Challenging Environments
Supervisor: Prof. Marcello Chiaberge, M. Martini, M. Ambrosio, G. Franchini
Skills: Basic Python (Mandatory) – ROS 2 (Recommended)
Code: 2025PIC02
Topic: Robot Localization
Description: LiDAR point cloud based odometry is one of the most studied family of algorithms in mobile robotics and autonomous driving. However, the actual performance of state of the art algorithms have been proved on driving datasets only with costly 3D sensor data. Service Robotics should face risky and hazardous environments such as tunnels, outdoor fields (vineyards), where the environments imposes hard contraints and offer few descriptors to localize the robot from data.
Objectives:
1) Benchmark most recent LiDAR odometry algorithms in these scenarios
2) Identidy a possible design for low-cost LiDAR sensor
3) Try to improve one of the studied approaches using low cost LiDAR sensor
Multirobot collaborative SLAM
Supervisor: Prof. Marcello Chiaberge, M. Martini, S. Primatesta
Skills: Basic Programming (Mandatory) – ROS 2 (Recommended)
Code: 2025PIC03
Topic: Multirobot SLAM
Description: Multirobot SLAM techniques aims at mapping the environments while localizing a fleet of robots. The thesis will explore distributed approaches to efficiently tackle a multirobot collaborative SLAM problem in a constrained environment of interest to be defined. Some examples could be multirobot SLAM in GPS denied settings (thick canopies and vegetation, tunnels, space…)
Efficient Person detection and tracking with 3D LiDAR Point Cloud
Supervisor: Prof. Marcello Chiaberge, M.Martini, A.Ostuni, F. Aisa
Skills: Basic Python (Mandatory) – ROS 2 (Recommended)
Code: 2025PIC04
Topic: Perception / Social Robotics
Description: Detecting and tracking people moving is a necessary perception ability of every service robot working alongside humans.Computer Vision methods are the most common solutions to address the problem, although often arising privacy problems in real case scenarios. Point clouds are rich data that can overcome this problem providing a complete scene screening, at the cost of additional computational effort. This thesis aims at investigating efficient solution to handle point clouds for detecting and tracking crowd moving.
Large Visual Language Models at the Edge
Supervisor: Prof. Marcello Chiaberge, S. Angarano, F. La Carpia
Skills: Basic Python (Mandatory) – Deep Learning (Recommended)
Code: 2025PIC05
Topic: Visual Language Models
Description: Study and benchmark most recent Large Visual Language Models on different computational hardware for inference at the edge. The activity of the student will support and integrate the ongoing work of PhD students on Edge AI for robotics
Large Visual Language Models for Robotics applications
Supervisor: Prof. Marcello Chiaberge, S. Angarano, F. Aisa, M. Martini
Skills: Basic Python (Mandatory) – ROS 2 (Recommended)
Code: 2025PIC06
Topic: Visual Language Models
Description: Explore most recent Large Visual Language Models, develop a ROS 2 integrated system for robot scene/command understanding based on VLMs, test the Human-Robot Interaction interface with navigation/grasping tasks where semantic commands are given.
Collaborative Manipulation Tasks for Robotic Harvesting in Agriculture
Supervisor: Prof. Marcello Chiaberge, S. Angarano, F. Aisa, M. Martini
Skills: Basic Python (Mandatory) – ROS 2, MoveIt (Recommended)
Code: 2025PIC07
Topic: Manipulation / Agriculture
Description: Develop the necessary manipulation and control expertise (ROS 2, MoveIt 2) to define a standard work framework for collaborative robotic harvesting.
VR/AR for robot learning from feedback and imitation
Supervisor: Prof. Marcello Chiaberge, B. Tuberga
Skills: Basic Python (Mandatory) – ROS 2, Unity (Recommended)
Code: 2025PIC08
Topic: VR / Robot Learning
Description: Integrate a robotic learning framework in a Virtual Reality / Augmented Reality environment to let the robot learn from the interaction with humans. The Virtual Reality environment can be used to generate task demonstration data to teach the robot how to solve complex tasks (navigation, grasping,…). The simulated scene should be develop in Unity, good programming skills or background experience with simulators is recommended (C, Python).
Panoptic Segmentation for Fruit Harvesting
Supervisor: Prof. Marcello Chiaberge, A. Navone
Skills: Basic Python (Mandatory) – Deep Learning (Recommended)
Code: 2025PIC09
Topic: Agriculture / Computer Vision
Description: One of the main difficulties of autonomous fruit harvesting is represented by the complex environment in which the robot is working. A visual understanding of the working scenario of the robot is fundamental to plan the movement of the robotic arm avoiding collision with solid obstacles (such as branches and wires) identifying the targets at the same time. Panoptic segmentation could represent a viable solution, identifying the different instances of the fruits, providing a semantic segmentation of the background and of the surrounding obstacles.
Thesis w/ Team RoboTO: weed detection and sprayer controller for the Field Robot Event Challenge
Supervisor: Prof. Marcello Chiaberge, A. Navone, G. Dara
Skills: Basic Python (Mandatory) – ROS 2, Deep Learning (Recommended)
Code: 2025PIC10
Topic: Agriculture / Computer Vision
Description: Support the development of the robotic solution for the International Field Robot Event Challenge 2025. PIC4SeR will partecipate to the challenge together with team RoboTO. The thesis work should focus on the development of a precise weed detection algorithm and spraying system to be placed and actuated on the robot used for the competition.
Thesis w/ Team RoboTO: Phisical and control design of a robotic arm for ground robots
Supervisor: Prof. Marcello Chiaberge, G. Dara
Skills: CAD design,Control Theory (Mandatory) – Python, ROS 2 (Recommended)
Code: 2025PIC12
Topic: System Design
Description: Support the development of the design and control of a robot arm for the Robomaster University Legue 2025 Engineering Challenge. The challenge involves two robots competing in an arena, where they must collect cubes and transport them to specific locations to score points. The project requires designing and controlling a robotic arm mounted on a mobile platform, enabling movement and precise manipulation during the match.
Thesis w/ Team RoboTO: Control barrier function for multi-robot navigation task
Supervisor: Prof. Marcello Chiaberge, M. Martini, A. Ostuni
Skills: Basic Python (Mandatory) – ROS 2, Control Theory (Recommended)
Code: 2025PIC11
Topic: Computer Vision/Navigation
Description: Support the development of the navigation solution for the Robomaster University Legue 2025 3v3 match. In this competition, two squads of three robots challenge each other in an arena. Two robot will be piloted and one will be completely autonomous. The autonomous robot should navigate freely inside the playground, reaching some key position in the arena and challenging the opponents, without crashing neither the allies nor the enemies. The aim of this thesis is to develop a navigation algorithm based on control barrier function, that takes all the competitions goals and strategy as constraint.