Robot Navigates Autonomously Using Vanishing Points
Writing by Evan Ackerman on Tuesday, 20 of January , 2009 at 2:16 am
A few months ago, we wrote about a robot helicopter that uses simple optical flow to navigate. The following video shows an iRobot Create using another conceptually simple technology to find its way around an office:
Basically, all the robot is doing is finding straight lines in the camera image, drawing those lines out, and then steering towards the point at which they converge. If there’s no convergence (if the lines are horizontal across the image) the robot knows that it’s looking at a wall and it needs to turn. It’s also able to recognize orange cones as landmarks.
Of course, this is only going to work somewhere with reliable geometry, like floors, walls, ceilings, windows, etc.. A completely blank webcam image of something like a white wall, or a complex image with no discernible straight lines like an outdoor scene, would most likely confuse this software. However, it’s designed for a specific environment and appears to work reliably and efficiently in that environment.
So, here’s a wild idea… Wouldn’t it be cool if someone could bundle together a bunch of these simple but environment-dependent navigation technologies into one package that a robot could dynamically access based on which proves to be more effective in whatever environment it finds itself in? Yeah, that would be cool. Someday, someday.
[ Hackszine ] VIA [ GoRobotics ]
Leave a comment
Category: Research
- Add this post to
- Del.icio.us -
- Digg
No comments yet.