Docking Station and Autonomous Parking for the Texas Robot

The Texas Robot continues to be a fun platform for experimenting with robot telepresence. One of the major difficulties with the original platform was charging. Although the Texas Robot could last an entire day on a single charge, the "pilot" always had to find someone to plug the robot in at the end of the day. Our hardware team did a great job solving this. They built a convenient docking station that's very easy to use -- you simply drive the robot backwards onto a platform, and a V-shaped bar helps guide the plug into place.

While this part of the operation is easy, we wondered if we could make it even more convenient. If the PR2 can plug itself in, why can't the Texas Robot dock itself? We want to improve the experience of robot telepresence, so that you can use robots like Texas as a platform for interacting with people, rather than spend time performing maintenance tasks. We haven't solved this problem entirely, but with the help of Victor Eruhimov and OpenCV, we got a step closer.

While visiting Willow Garage last December, OpenCV developer Victor Eruhimov combined OpenCV and ROS to develop a vision-based prototype of autonomous parking. This prototype uses the color camera on the Texas Robot to find the docking station, which has a grid of color circles on it. A color filter is applied to the camera image, and a search is done for the pattern of circles. Once the docking station is located, the Texas Robot can be driven into place -- the clever design of the docking station makes it just as easy for an autonomous robot to use as well. This approach still requires that the driver position Texas so that it can see the docking station, but hopefully we can address that with future improvements.