Open Positions - Master Thesis

Application Open: 02/11/2026

Application Deadline: 07/12/2026 

Degradation-Aware Multimodal Localization for GNSS-Denied Field Robots

Supervisor:  Prof. Marcello Chiaberge, M. Martini, G.Audrito, M.Ambrosio

Skills: Mandatory: Python/C++.  Recommended: ROS 2, GTSAM, SLAM concepts

Code: 2026PIC01

Topic: Field Robotics, Localization, Sensor Fusion

Description: The project aims to develop a robust localization system for robots operating in dark, confined, hazardous, or GNSS-denied environments. LiDAR, IMU, and visual measurements will be integrated through a factor graph. The system will detect sensor degradation and adapt measurement weights according to estimated uncertainty and directional observability. Activities include benchmarking of state-of-the-art methods on available simulators, and the collection of experimental data on specific fields sites for experimental validation on a mobile robot.

Gaussian Splatting for 3D Robotic Mapping

Supervisor:  Prof. Marcello Chiaberge, M. Martini, G.Audrito, M.Ambrosio

Skills: Mandatory: Python/C++. Recommended: ROS 2, PyTorch, Computer vision

Code: 2026PIC02

Topic: 3D Mapping, Spatial AI

Description: The project aims to study and develop a dense 3D mapping system of the environment by combining LiDAR geometry with Gaussian Splatting. LiDAR measurements will constrain Gaussian initialization and optimization to improve metric accuracy, while camera images will provide appearance information. Evaluation will be conducted first with different state-of-the-art methods. The resulting maps will be evaluated in terms of geometric accuracy, rendering quality, memory use, and runtime on public datasets and robot data.

Multi-temporal Fruit Mapping in Vineyards and Orchards

Supervisor:  Prof. Marcello Chiaberge, M. Martini, G.Audrito, A. Navone, M.Ambrosio

Skills: Mandatory: Python, computer vision. Recommended: PyTorch, ROS 2

Code: 2026PIC03

Topic:Agri-Robotics, Computer Vision

Description: Precise fruit counting and monitoring are fundamental activities for yield estimation and preservation in agriculture. This master’s thesis will investigate robotics and AI methods for generating geo-referenced fruit maps from RGB-D/IMU/GPS data collected by a rover in vineyards and orchards. Starting from an existing factor graph-based semantic SLAM implementation, the project will adapt and extend the approach to different field environments and repeated surveys over time. The work will address fruit detection, localization, and mapping across different sessions to track changes in fruit distribution and development. Finally, the methods will be evaluated on field data to assess mapping accuracy, temporal consistency, and robustness across different agricultural settings.

Robust Visual Place Recognition in Vineyards

Supervisor:  Prof. Marcello Chiaberge, M. Martini, A. Navone, G. Audrito

Skills: Mandatory: Python, computer vision. Recommended: PyTorch, ROS 2

Code: 2026PIC04

Topic: Agri-Robotics, SLAM, Computer Vision

Description: The project aims to develop a visual place recognition system for reliable robot localization and loop closure in vineyards. The method should address perceptual aliasing caused by repetitive rows and appearance changes due to season, illumination, weather, and viewpoint. The work will investigate efficient global and local image descriptors, temporal or geometric verification, and lightweight model deployment. The system will be evaluated on public and newly collected vineyard sequences using retrieval accuracy, localization robustness, runtime, and memory consumption.

Human Semantic Cue Recognition for Intention and Cooperation Prediction in Robot Navigation

Supervisor:  Prof. Marcello Chiaberge, M. Martini, S. Trepella, F. La Carpia

Skills: Mandatory: Python, computer vision. Recommended: PyTorch, ROS 2

Code: 2026PIC05

Topic: Human-Aware Robotics, Computer Vision

Description: The project aims to develop an RGB-based perception system that recognizes human semantic cues relevant to robot navigation, including gaze direction, body orientation, gestures, motion, and group interactions. These cues will be used to predict pedestrian intentions and willingness to cooperate with a robot, such as yielding or maintaining priority. The work includes integrating existing pose and tracking models, developing a temporal prediction method, and evaluating its contribution to socially aware navigation using public datasets and robot experiments.

Human-Robot Interaction Benchmarking in VR

Supervisor:  Prof. Marcello Chiaberge, A. Ferri, M. Dipaola

Skills: Mandatory: CAD, URDF, ROS 2. Recommended: Unity, Python, C#

Code: 2026PIC06

Topic: Human-Robot Interaction, Robotic Modeling and Simulation

Description: The project aims to develop a digital twin of a novel mobile manipulator designed for human-robot interaction, enabling its simulation and testing in a Unity virtual reality environment. The platform will be used to conduct human-robot interaction experiments involving the use of a VR headset and evaluate the robot’s behavior using social metrics, providing a controlled and repeatable framework for assessing its performance.

Localization and Planning for Lunar Exploration with Autonomous Rover

Supervisor:  Prof. Marcello Chiaberge, M. Dipaola, G. Audrito, M. Ambrosio, M. Martini

Skills: Mandatory: C++, Basic Python. Recommended: ROS 2

Code: 2026PIC07

Topic: Space Robotics

Description: The project aims to develop a planning and localization system 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 project should be developed in accurate simulations of the lunar surface and possibly tested in a real testing facility

4D Scene Graphs for Human-aware Navigation

Supervisor:  Prof. Marcello Chiaberge, M. Martini, F. La Carpia, A. Ostuni, S. Trepella

Skills: Mandatory: Python. Recommended: PyTorch, ROS 2, computer vision

Code: 2026PIC08

Topic: Human-Aware Robotics

Description: This master’s thesis will investigate the use of 4D scene graphs spatio-temporal representation of environments to recognize hum\an activities and social contexts for human-aware robot navigation. The project will begin with a review of state-of-the-art methods for scene graph construction and application in human context recognition. Promising approaches will then be selected, adapted, and evaluated in relevant scenarios. Finally, the methods will be integrated into a robotic platform and tested to assess their ability to support safe, socially appropriate navigation in shared human–robot environments.

Design and Development of an Autonomous Shpherical Payload for Lunar Exploration

Supervisor:  Prof. Marcello Chiaberge, M. Dipaola

Skills: Mandatory: SolidWorks/Fusion 360, C++.  Recomanded: Basic Electronics. Extra: PCB Design.

Code: 2026PIC09

Topic: Space Robotics, Embedded Systems, Low-Power Electronics

Description: The thesis aims to design and develop a depoyable spherical payload for lunar exploration, integrating mechanical design, embedded electronic system, solar panels, battery storage and wireless communication. The project will focus on mechanical robustness, energy autonomy, communication reliability, and system integration. The work will culminate in a functional prototype validated through simulations and experimental testing.

UWB robust multi-robot localization

Supervisor:  Prof. Marcello Chiaberge, G. Audrito, M. Dipaola

Skills: Mandatory: Python. Recommended: ROS 2

Code: 2026PIC10

Topic: Multi-robot Localization, Field Robotics

Description: This master’s thesis will investigate robust multi-robot localization using ultra-wideband (UWB) sensors. The project will begin with a review of state-of-the-art methods and benchmark selected approaches on an existing indoor dataset, assessing localization accuracy and robustness under challenging sensing conditions. Improvements of staarting methods can be proposed and developed along the thesis project. The dataset will then be extended through outdoor experiments involving heterogeneous teams of drones and rovers. Finally, the selected methods will be evaluated on the expanded dataset to assess their performance across different environments and robot configurations.