Making robots huggable and safe | Gordon Cheng | TEDxTUM

In this talk

1. 0:23

   Why robots need skin

   Human skin helps us handle objects and interact with our surroundings. Cheng argues that sensors across a robot’s body could make industrial and home robots safer, while also enabling new uses in health care, sports, and entertainment.

2. 2:13

   Building flexible, durable sensors

   Cheng describes sensor cells that detect temperature, proximity, pressure, and vibration. The team explored how to distribute them across a large surface so the skin can adapt to shapes and withstand impacts, cuts, and bending.

3. 5:27

   Mapping the skin and learning movement

   The team developed algorithms to reconstruct the robot’s 3D surface and locate individual sensor cells. They also taught the robot to associate sensory input with movement, demonstrating control of an arm through gentle contact, including with a feather.

4. 7:52

   Processing touch efficiently

   Human skin has about five million receptors, and conventional computing approaches would require too much processing for a robot covered in sensors. Cheng describes using event based processing inspired by biology, where information is sent when pressure or another sensation changes significantly.

5. 9:21

   Using skin for balance and safe contact

   The team covered a humanoid robot with about one square meter of sensor skin. Skin based pressure estimates helped it balance and handle uneven ground, such as plants or pebbles; Cheng also demonstrated whole body interaction designed to let the robot hug a person safely.