Manipulation Researcher

Giovanni Franzese, Ph.D.

Giovanni is a Postdoctoral researcher in Robotics at TU Delft in the Netherlands. He obtained his Ph.D. in 2024 from TU Delft with the thesis named Uncertainty-aware Interactive Imitation Learning for Robot Manipulation. During his Ph.D. he focused on estimation and rejection of uncertainty in robot learning and algorithm design to allow robot to fastly learn from human demonstration and corrections.

About Giovanni

He is passionate about developing algorithms that enable robots to learn complex manipulation tasks from human demonstration and feedback. With experience in programming real robot controllers and training machine learning models, he has deployed controllers for Franka Emika robots in universities worldwide, supporting research for scientists and students. He advocates open science and he is committed for tackling the complex problem of robot manipulation to improve the quality of life of people.


Work experience

Postdoctoral Researcher
Delft University of Technology, Delft, Netherlands
Sep. 2023 - Present

Mentored by Cosimo Della Santina. Part of the NXTGEN high-tech agri-food Project.
Managed the Franka Emika lab of Cognitive Robotics at TU Delft, exploring learning-based methods for low-cost, soft, and deformable hardware, with a focus on generalizing manipulation skills in tomato plant de-leafing.

Team Leader, PLATOnics Team
Delft University of Technology, Delft, Netherlands
2023 - 2024

Led PLATOnics team for the EUrobin Manipulation Skill Versatility Challenge 2024 and the Robothon Manipulation Challenge 2023.
Developed a solution based on human demonstrations, awarded as the best versatility solution at IROS 2024.

Visiting Ph.D. in Statistical Machine Learning
University College London (UCL), London, United Kingdom
Sep. 2022 - Feb 2023

Mentored by Marc Deisenroth. Focused on variational inference learning methods for calibrated classification on high-dimensional inputs, including images and graphs.

Ph.D. in Interactive Imitation Learning for Robotics
Delft University of Technology, Delft, Netherlands
June 2019 - June 2023

Thesis: "Uncertainty-aware Interactive Imitation Learning for Robot Manipulation." Mentored by Jens Kober and Luka Peternel. Supported by ERC grant TERI (Teaching Robot Interactively).

Lecturer for MSc Course on Intelligent Control Systems
Delft University of Technology, Delft, Netherlands
2022-24

Lectured on Gaussian Processes for Robotics and Control.

Skills

A snapshot of the skills and expertise I bring to robotics and machine learning.

Robot Control

Variable Impedance Control for Single and Bimanual Manipulation. Safe human-robot interaction.

Robot Learning

Low-level and high-level skill learning from human interactive demonstrations.

Machine Learning

Experience in Gaussian Process, Epistemic Uncertainty Estimation, Calibration, and Variational Inference.

Let's teach robots together

Ready to make robot do great things? Let's work together to make it happen.