Giacomo Spigler
I am a tenured assistant professor at Tilburg University and co-director of the AI for Robotics Lab (AIR-Lab).
I work on robot learning, mainly imitation learning, reinforcement learning and active perception, with a focus on humanoid robots. I develop most of my own research code, and I build robots on the side.
Selected Work
- TAVIS (2026). A benchmark for active vision in imitation learning, where the policy also decides where the robot looks. Two humanoid torsos in IsaacLab, 2,200 VR-teleoperated demonstrations, Diffusion Policy and π0 (VLA) baselines, in- and out-of-distribution evaluation splits, and a metric to measure gaze-hand anticipation.
arXiv · website · code · data and checkpoints
- POUCH (2026). An object-centric memory for visuomotor policies that tracks objects through occlusion, including objects carried inside moving containers, and conditions a Diffusion Policy through an object-centric set transformer. Built on GroundingDINO and SAM2, and evaluated in simulation and on Reachy 2, with the University of Modena and Reggio Emilia.
code
- Reachy 2 at AIR-Lab. Our humanoid platform, with a custom-made Quest 3 teleoperation interface that records directly into LeRobot datasets, on the real robot and in simulation.
code · video
- Proximal Policy Distillation (TMLR 2025). A policy distillation method built on PPO, in which the student keeps learning from reward while being distilled and can outperform its teacher. Implementation for Stable-Baselines3.
paper · code
- Tilburg Hand. A low-cost dexterous robot hand I designed and built, presented at the ICRA 2023 workshop on open-source manipulation ecosystems.
Tilburg Robotics
- Flatland-RL. Co-author of the multi-agent RL benchmark for train scheduling used in the NeurIPS 2020 Flatland competition, built at AIcrowd with SBB.
arXiv
Current Position
- (January 2022 - current) Tenured Assistant Professor in Artificial Intelligence at the Department of Intelligent Systems of Tilburg University.
Co-director of the AI for Robotics Lab (AIR-Lab) and of the MindLabs Robotics & Avatars Lab.
- Research fellow at SIKS; mentor at LeadTheFuture.
Invited Talks
- Building the ChatGPT of Robotics Requires Physical Intelligence, plenary talk at the International Conference of Physics Students (ICPS), Rabat, Morocco. August 2026.
- Deep Reinforcement Learning in Robotics: Bridging the Gap between Simulation and Real-world Applications, "Contributions to Control Systems" seminar series, Institute of Automatic Control, RWTH Aachen University, Germany. May 2024.
- AI revolution & Robotics: towards highly-autonomous agricultural robots, symposium "Agricultural Resources, Leadership and Artificial Intelligence", Italian Embassy, Den Haag, The Netherlands. October 2023.
- End-to-end robot learning and the Tilburg Hand, TAISIG Talks 15, Tilburg, The Netherlands. August 2022.
Service
- Area Chair, CoRL 2026. Top Reviewer, NeurIPS 2026.
- Organizer of BeNeRL 2026, the Belgium-Netherlands Workshop on Reinforcement Learning, with Marija Popović.
- Organizer of the AI for Robotics Workshop 2024, with Murat Kirtay.