RoboBoom
University of Toronto Robotics Institute· December 8, 2020· 59:15

Seminar: Dieter Fox at UofT Robotics Institute

University of Washington professor and NVIDIA Senior Director of Robotics Research Dieter Fox presents his lab’s work on robot manipulation at the University of Toronto Robotics Institute seminar.

In this talk

  1. 1:51

    Building a kitchen manipulation system

    Fox describes research aimed at robots that can handle complex tasks in unfamiliar settings, including homes, hospitals, and industrial workplaces. His team built a kitchen test environment to study challenges shared across these domains, using a mobile Franka Panda arm and overhead depth cameras.

  2. 9:38

    Estimating object poses and environment state

    Fox presents Pose RBPF, a particle-filtering approach that represents uncertainty across an object’s full 6D pose, including ambiguous orientations caused by symmetry or limited views. He also describes DART, which tracks articulated objects such as robot arms, people, cabinets, and drawers from depth data, assuming their models are known.

  3. 28:31

    Reactive control and task planning

    Riemannian Motion Policies combine local obstacle avoidance with motion toward a goal, producing fast reactions in the robot’s joint space. Fox shows how a planner can adapt when a person changes the scene, such as closing a drawer the robot needs open, and explains that these systems rely on models of the environment and objects.

  4. 36:34

    Grasping unknown objects and using simulation

    Fox describes 6-DoF GraspNet, trained on simulated grasps, which proposes and scores grasps from an object point cloud. In clutter, the system can identify an obstructing object, move it, and retry the target grasp. He also discusses simulation for training and testing, while noting that creating accurate digital environments remains labor-intensive.

Briefs summarise talks and interviews; timestamps open the video at that moment. Claims in a talk are the speaker's own.

More talks