Indoor Mobile Robot Localization

Posted 30 Apr 2007 at 18:14 UTC by steve

The Robotics Institute released a report recently titled Parrots: A Range Measuring Sensor Network (PDF format) that describes a wireless sensor network that can be used for localization by indoor mobile robots. The network consists of nodes, called Parrots, that measure the range to other nodes using ultrasonic sensors. Each node includes a 16bit microcontroller, a wireless link, and a sensor board with 4 ultrasonic sensor arrays. When lots of Parrots are placed in an area, they automatically form a multi-hop ad-hoc wireless network. The robot carries a Parrot node as well, allowing the network to accurately map the location of the robot as it moves through the area. This research was done by the CMU Field Robotics Center.

Rather impractical, posted 1 May 2007 at 10:02 UTC by motters » (Master)

This may work, but it still seems like a very unsatisfactory solution to me. Having to litter the area with ultrasonic transmitters doesn't seem terribly practical, although its possible that if they could be manufactured cheaply enough this might be a way of having a relatively inexpensive robot navigation system in the home.

To do robot navigation reliably without ad-hoc modifications to the environment you really need cameras or lasers (or both). At present lasers are out of the equation because they're still too expensive for anything other than research or high end industrial applications. Cameras however are much more economical, but require the processing power of a modern PC to meaningfully interpret the data. So you either need a PC onboard the robot, or some hardware which can transmit camera data over a wifi network to a desktop PC which can then do the number crunching.

possibly usefull, for a slightly different purpose, posted 1 May 2007 at 12:52 UTC by MDude » (Journeyer)

Although having to put sensors everywhere dosn't seem very usefull, if the robots themselves carry them then I think it can a good way to control swarm robots. As for lasers, I'm not sure about the expensiveness of them. What kind of laser and other materials are requred for a good range finder?

laser ranger principle of operation, posted 2 May 2007 at 21:51 UTC by motters » (Master)

For your typical laser ranger a laser beam fired in the vertical axis hits a spinning mirror or prism which redirects the beam through 90 degrees and out at the current angle in the horizontal plane. The beam then shoots out, and is reflected back from an object hitting the mirror again and being redirected back up to a sensor. Knowing the speed of light (it's fast!) and the total time of flight the system can work out the range to the object surface. All this has to be done quickly to be of any practical use on a robot moving at speed.

The cost of scanning laser rangefinders does seem to be falling, but not very fast. If some company were to come out with a device costing $50 I think that would change the robotics landscape, but it doesn't look as if this is likely to happen any time soon. Currently scanning laser rangefinders cost around $2000.