Robot Vision and Learning Lab

Welcome

Welcome to the Robot Vision and Learning (RVL) lab. We are part of the Computer Science department at the University of Toronto, the MCS department at UTM, the UofT Robotics Institute, and Vector Institute. The group is led by Prof. Florian Shkurti, and consists of students with backgrounds in robotics, machine learning, computer vision, engineering, control theory, or physics. We develop methods that enable robots to perceive, reason, and act effectively and safely, particularly in dynamic environments and alongside humans. Application areas include robotic manipulation, chemistry/biology lab automation, and field robotics for environmental monitoring.

Research Themes

• reasoning and learning for robotics
• safe robot learning
• 3D vision & generative models for simulation

Application Areas

• robotics for chemistry and biology lab automation
• field robotics for environmental monitoring

News

• Sep 2026: Qi Chen's paper on using foundation models to inform acquisition functions for molecular discovery was accepted at NeurIPS'26. This is a collaboration with Fabio Ramos and Alán Aspuru-Guzik. Congratulations, Qi!
• Aug 2026: Three papers accepted at CoRL'26: one on a fast, model-free diversity metric for robotics datasets, selected as an oral, another on torque adaptation for robust motion transfer in manipulation, selected as a spotlight, and a third on what you need for compositional generalization in diffusion planning. Congratulations to Sreevardhan Sirigiri, Nathan de Lara, Dongwon Son, Quentin Clark, and the rest of the co-authors!
• May 2026: Nathan de Lara's paper SMAC on score-matched actor-critics for robust offline-to-online transfer was accepted at ICML'26. Congratulations, Nathan!
• Jan 2026: Our paper RoboCulture on modular robotic plate-based cell automation was published in Cell Reports Physical Science. Congratulations to Kevin Angers, Kourosh Darvish, and our collaborators in Milica Radisic's and Alán Aspuru-Guzik's groups!
• Nov 2025: Wei-Cheng's paper on leveraging video models for policy evaluation in robotic manipulation is out on arxiv.
• Sep 2025: Two papers accepted at NeurIPS'25, one on off-policy evaluation with diffusion guidance and another on multi-task failure detection in VLAs. Both papers were collaborations with TRI. Congratulations to Hossein Goli, Mike Gimelfarb, Qiao Gu, Yuanliang Ju, and the rest of the team!
• Sep 2025: One paper accepted at Nature Computational Science on digital twins for chemistry labs and another paper (viewpoint) accepted to Science Robotics. Congratulations to Kourosh Darvish and all the collaborators!
• Sep 2025: New PhD students Hossein Goli, Brandon Huang, Zhengyang Liang, Yuanliang Ju and new postdoc Giseung Park join the lab. Welcome to RVL!
• Aug 2025: Two papers accepted at CoRL'25, one on learning policies for generalizable placement and another on test-time adaptation of VLMs for Bayesian search. Congratulations to Yuchi Zhao, Derek Tan, and the rest of the co-authors!
• Jun 2025: Ritvik Singh's paper Synthetica on large-scale synthetic data generation for robot perception, a collaboration with NVIDIA, was accepted at IROS'25. Congratulations to Ritvik and the co-authors!