PhD calls 2025
Fully funded Phd positions are available in the
Humanoid Sensing and Perception Istituto Italiano di Tecnologia.
The positions are available through the PhD course of
Ph.D. program of national interest in Robotics and Intelligent Machines (DRIM), curriculum on
Hostile and Unstructured Environments.
For more details on the research topic read below.
Applications deadline: July 9th 2025 at 12 (noon – CEST).
Application instructions: refer to the
official DRIM admission page.
Important: to apply, you should select the
Hostile and Unstructured Environments curriculum, and the Research theme #3
Learning Adaptive Robotic Behavior in Unstructured Environments.
Prospective candidates are invited to get in touch with Lorenzo Natale (name.surname@iit.it)
for further details.
Learning Adaptive Robotic Behavior in Unstructured Environments
Description:
The deployment of robots in unstructured environments remains a central challenge in robotics and artificial intelligence. Recent advances — particularly the emergence of foundation models (LLMs and VLMs) – have enabled promising forms of robot programming through natural language and task planning. However, their utility is mostly limited to highlevel instruction following, and an open problem remains how to ground pre-trained model to the robot action space. Imitation Learning (IL) and Reinforcement Learning (RL), on the other hand, have been widely used to train robots for complex tasks, yet they come with significant limitations: IL relies on human teleoperation and is labor-intensive, while RL demands extensive simulation and computational resources. Both approaches scale poorly and struggle to support autonomous adaptation in real-world deployment.
This PhD project aims to investigate learning frameworks that overcome these limitations and enable scalable, adaptive behavior in humanoid robots with a focus on generalization and physical interaction.
Several research directions may be pursued depending on the candidate’s interest and background, including:
- Motion retargeting while inferring physical interactions from egocentric videos;
- Blending IL and RL techniques to enable robust policy transfer across embodiments and environments;
- Visuo-tactile sensor integration to solve contact-rich tasks with minimal supervision;
- Physics-aware particle based scene representation and simulation;
- Task learning and grounding through natural language interaction, including learning and adaptation of foundational multi-modal models.
The goal is to develop methods that enable robots to learn efficiently, adapt autonomously, and generalize across environments, embodiments, and tasks.
Requirements:
The ideal candidate would have a degree in Computer Science, Engineering or related disciplines, with a background in Robotics, Computer Vision and/or Machine Learning. They would also be highly motivated to work on robotic platforms and have computer programming skills.
References:
[1] Ceola, F., Rosasco, L., and Natale, L., RESPRECT: Speeding-up Multi-fingered Grasping with Residual Reinforcement Learning, IEEE Robotics & Automation Letters, vol. 9, no. 4, 2024
[2] Puang, E.Y., Ceola, F., Pasquale, G. and Natale, L. PCHands: PCA-based Hand Pose Synergy Representation on Manipulators with N-DoF, under submission 2025.
[3] Vasile, F., Qiu R., Natale, L., Wang X., CollidingGS: Gaussian-Augmented Physics Simulation and Identification with Complex Colliders, under submission 2025.
Contacts: Lorenzo Natale (name.surname@iit.it)