Neato's triangulating laser scanner
Most current mobile robots use a laser scanner for building maps and these things cost serious money. Therefore, when Neato built a laser scanner into a vacuum for <400$ and claimed the scanner cost less than <30$ many people became interested very fast. Hizook had an interesting blog article about Neato's scanner, linking to a 2008 ICRA paper by the Neato people and none other than Kurt Konolige, now at Willow garage.
Well, the principle is easy to explain: Instead of measuring time-of-flight, which requires expensive circuitry, the device measures deflection from the center -- which is also sensitive to distance.
However, upon reading the paper, one thing caught my eye: They support up to 4KHz on the camera. That means taking 4000 images per second, whereas a conventional camera only takes 30. The reason is simple: To cover the whole 360 degrees, the laser is rotated and an image is taken a steps of 1 degree. The sensor does so at 10Hz, which makes 3600 readouts/s.
Furthermore, this is not a simple line-scanner -- due to inaccuracies in the laser and the mounting, the need 5 lines above and below the center scanline. To read these out at the given rate, on-chip line binning is used (for distance, you don't care about the y offset).
Now, none of that is rocket science, but it probably explains why we haven't seen cheap knock-offs with normal webcams. You need to be familiar with using CMOS sensors directly to do this kind of sensor. It is also just one of the many practical issue the Neato people had to overcome and the rest of the paper also makes quite an interesting read. Recommended :-)
Well, the principle is easy to explain: Instead of measuring time-of-flight, which requires expensive circuitry, the device measures deflection from the center -- which is also sensitive to distance.
However, upon reading the paper, one thing caught my eye: They support up to 4KHz on the camera. That means taking 4000 images per second, whereas a conventional camera only takes 30. The reason is simple: To cover the whole 360 degrees, the laser is rotated and an image is taken a steps of 1 degree. The sensor does so at 10Hz, which makes 3600 readouts/s.
Furthermore, this is not a simple line-scanner -- due to inaccuracies in the laser and the mounting, the need 5 lines above and below the center scanline. To read these out at the given rate, on-chip line binning is used (for distance, you don't care about the y offset).
Now, none of that is rocket science, but it probably explains why we haven't seen cheap knock-offs with normal webcams. You need to be familiar with using CMOS sensors directly to do this kind of sensor. It is also just one of the many practical issue the Neato people had to overcome and the rest of the paper also makes quite an interesting read. Recommended :-)
- ingo's blog
- Log in to post comments
Recent Photos
