Interview with Ales Ude (JSI): Humanoid robots, imitation learning and torque control

In this talk

1. 0:14

   From perception to imitation learning

   Ude says he came to robotics by chance and grew interested in its real-world complexity. His early work focused on robot vision and interpreting human motion. He describes imitation learning as transferring human movements to robots and improving their behavior through practice.

2. 1:40

   Learning to generate new behaviors

   Ude says robots need to do more than repeat observed movements because they encounter new situations. His team collects libraries of human movements, transfers them to humanoids, and uses the data to generate new behaviors. He sees robots as a way to shift people from simple manual work toward other jobs and services.

3. 3:22

   Torque control and more flexible production

   Ude says the team chose TALOS because torque sensors in most of its joints support torque control. Sensing contact forces can help a robot respond compliantly and interact with people and objects. He says more versatile robots could also make automation practical for smaller production runs, where programming repetitive tasks is harder to justify.

4. 5:14

   Robots in services and at home

   Ude expects robots to become more useful in settings such as hospitals and shops, though current examples are not very advanced. He says home robots that perform many different daily tasks are more than 10 years away. Even limited support for older people could help them stay at home longer.

5. 6:33

   Studying human-robot contact

   Ude says TALOS lets the team study compliant humanoids through torque control. Their research examines how robots can interact with people and use contact with the environment. He hopes robots will increasingly support people with daily tasks.