Florian Shkurti

Assistant Professor, Computer Science

I direct the Robot Vision and Learning (RVL) lab at the University of Toronto. I am part of the Robotics Group at the Department of Computer Science. My research centers around robotics and spans machine learning, perception, planning and control. My group and I develop methods that enable robots to perceive, reason, and act effectively and safely, particularly in dynamic environments and alongside humans. I want to enable people to effectively specify the desired task and behavior of their robots, without requiring them to write code. I also strive to provide safety guarantees during the robot learning process and beyond, during deployment. Application areas of my research include field robotics for environmental monitoring, visual navigation for autonomous vehicles, and mobile manipulation. I am interested in the following topics:

Autonomous robots for environmental monitoring and scientific exploration

• Learning visual attention operators for exploration and navigation
• Active sensing and information gathering
• Informed visual search
• Visual navigation
• Video and sensor data summarization
• Field robotics

Machine learning for planning and control

• Imitation learning
• Inverse reinforcement learning
• Planning and control under uncertainty
• Learning-based task and motion planning
• Differentiable physics and rendering simulators
• Likelihood-free inference
• Continual learning
• Shared autonomy

Safety guarantees during robot learning and deployment

• Safe exploration in reinforcement learning
• Physically-realizable adversarial simulation scenarios
• Data-driven sensor simulation
• Rare event generation
• Risk estimation

One of the applications of my research has been to enable environmental monitoring by robots that autonomously operate in the field and collect useful data for environmental scientists. Human-robot collaboration, learning from a few demonstrations, and robust navigation in 3D are critical in these outdoor domains. I have also worked on vision-based simultaneous mapping, localization, and navigation for ground, aerial, and underwater robots. Another application of our research has been in autonomous and vision-based robot manipulation for chemistry lab automation.