If you've ever gotten stuck in a t-shirt, this robot is for you.
For most people, putting on a t-shirt isn't a chore, but researchers at Nara Institute of Science and Technology in Japan have identified this important task as particularly difficult for the elderly or disabled with limited arm movement.
A cross-laboratory team led by Tomohiro Shibata and Takamitsu Matsubara developed a two-arm robot to slide a shirt over and onto a person's head and torso. Since a person's neck or arms may not be in the exact same position each time, a scripted movement could potentially cause distress.
Enter the team's reinforcement learning approach. Just like a child learning through experience, the robot is taught once how to clothe a human user, and then is given several attempts to put the shirt on by itself. The success is measured with motion capture system at the end of each trial, which lasts about 10 seconds.
In the video, we see that after three learning trials, the robot has learned the trajectory to place the shirt on without any trouble. According to Shibata, Japanese reporters who tested out the system gave it a thumbs up, saying it makes it easy to put on.
For now, the system has been tested with only a couple of different t-shirts and with several subjects including a few patients. The next step, says Shibata, is to try out the system with more subjects and patients, and with different t-shirts. "Our approach could be applied to other types of important clothing tasks such as pulling up/down pants."
Shibata, Matsubara, and their colleagues T. Tamei and A. Rai presented a paper, "Reinforcement Learning of Clothing Assistance with a Dual-arm Robot," describing their results at the 2011 IEEE-RAS International Conference on Humanoid Robots held in Bled, Slovenia, last week.
Angelica Lim is a graduate student at the Okuno and Ogata Speech Media Processing Group at Kyoto University, Japan.