I am a PhD researcher in Automotive Engineering for Intelligent Mobility at the University of Bologna, working at the intersection of artificial intelligence, autonomous systems, computer vision, simulation and robotics.

My work focuses on building systems that can perceive complex environments, estimate state, make decisions and act safely, with a strong emphasis on reproducible simulation, measurable validation and the transition from virtual environments to real platforms.

I currently work across three tightly connected research areas:

• Autonomous & intelligent vehicles — perception, digital twins, decision-making, safety and simulation.
• Computer vision & human perception — semantic scene understanding, eye-gaze analysis and multimodal driving datasets.
• Underwater robotics — BlueROV2-class autonomous navigation, obstacle avoidance, multi-robot coordination and HoloOcean/ROS 2 simulation.

Research theme: AI-driven simulation for perception and decision-making in autonomous systems.

Camera-based underwater obstacle avoidance with ROS 2, HoloOcean and Unreal Engine.
A closed-loop perception-to-control pipeline using RGB vision, YOLO detections, estimated odometry and a local avoidance planner. The system keeps simulator ground truth outside the runtime navigation loop and uses it only for validation.

ROS 2 · HoloOcean 2.3 · Unreal Engine · YOLO · Computer Vision · Autonomous Navigation

Marine Race Arena for autonomous underwater gate racing and multi-ROV research.
A configurable benchmark architecture separating participant autonomy from referee truth, with onboard-only controllers, acoustic beacons, current profiles, obstacles, coordination and reproducible HoloOcean validation.

Underwater Robotics · BlueROV2 · Autonomous Racing · Multi-Agent Systems · Simulation

Semantic scene understanding and eye-gaze analysis for Meta Project Aria Gen 2 driving recordings.
A fully local pipeline combining two independent segmentation approaches, fisheye rectification, synchronized gaze projection, canonical semantic classes, comparison metrics and visual overlays.

Project Aria Gen 2 · Grounding DINO · SAM 2.1 · Mask2Former · Eye Gaze · Semantic Segmentation

Hybrid-automaton digital twin and CARLA pedestrian protection system.
Combines vision-based pedestrian detection with formal braking logic and a digital-twin architecture to evaluate pedestrian AEB scenarios in simulation.

CARLA · YOLOv8 · Digital Twin · Hybrid Automata · MQTT · Autonomous Driving

Research work around Assetto Corsa, vehicle dynamics and intelligent driving simulation, connected to my broader work on autonomous mobility and data-driven vehicle-state estimation.

Automotive · Simulation · Vehicle Dynamics · AI

AI & perception: PyTorch, TensorFlow, Scikit-learn, OpenCV, YOLO, Grounding DINO, SAM 2, Mask2Former
Robotics & autonomy: ROS 2, BlueROV2, HoloOcean, MAVLink-oriented architectures, sensor fusion, local planning
Simulation: CARLA, HoloOcean, Unreal Engine, Assetto Corsa
Software: Python, C, C++, C#, Java, Kotlin, Swift, JavaScript, PHP, MATLAB
Engineering: Git, Docker, Linux, Windows, Conda, CI/testing, reproducible experiment pipelines

Real problem
    ↓
Simulation / data acquisition
    ↓
Perception & state estimation
    ↓
Decision-making / control
    ↓
Closed-loop evaluation
    ↓
Metrics, validation and reproducibility
    ↓
Transfer toward real systems

I am particularly interested in research where AI is evaluated as part of a complete autonomous system, rather than only as an isolated model.

I am currently expanding my work on:

• robust perception and decision-making for autonomous systems;
• underwater autonomy using realistic sensor-driven simulation;
• reinforcement learning and hybrid control for autonomous navigation;
• semantic driving-scene understanding with wearable sensing and eye tracking;
• reproducible simulation-to-real research pipelines.