Gianluca.
Computer Engineering graduate specialised in robotics, autonomous systems, and intelligent perception.
Hi, I'm
About Me
My focus is on robotics, spanning classical and modern control, real-time perception, and AI-driven task planning. I also enjoy stepping into new territory and tackling problems I haven’t faced before, whatever the domain.
I like to understand things from the ground up and find solutions that actually work, driven by curiosity and a hands-on approach to engineering.
Projects
ROS2
Python
MoveIt2
LLM / Gemini
Intel RealSense
PCL
libfranka
Active Perception and Language-Conditioned Task Planning for Robotic Manipulation
My Master’s thesis. The goal was to let a robot arm understand plain language instructions
and act on them autonomously. A multimodal LLM breaks the instruction into a sequence of
executable actions, while an active perception module keeps scanning the scene until it
builds a reliable model of each target object before attempting a grasp. Vision is handled
by an Intel RealSense D435i mounted on the end-effector, deployed on a real
Franka Emika Panda arm.
ROS2
Python
YOLO
Gazebo
Franka Panda
Kinect
Human-Robot Collaborative Cell
A collaborative robotic cell where the robot tracks the operator’s skeleton in real time
using a Kinect RGB-D camera and the YOLOv8-pose model, adapting its motion through
an impedance controller so it reacts softly to contact. When the operator enters the
robot’s proximity, the arm stops and switches to compliance mode, allowing the operator
to guide it by hand. A Kalman filter handles partial occlusions of body keypoints between
frames. The full pipeline was first validated in Gazebo before running on the real
Franka Emika Panda arm.