Servo - February 2009

Our resident expert on all things robotic is merely an email away.

roboto@servomagazine.com

Tap into the sum of allhumanknowledge and get your questions answered here! From software algorithms to material selection, Mr. Roboto strives to meet you where you are — and what more would you expect from a complex service droid?

by

Dennis Clark

Reader Responses

Let it not be said that Mr. Roboto suffers the sin of hubris. Some readers responded back to me about their ideas on how to deal with a December question dealing with non-GPS robot navigation on the cheap. There are some good ideas here, to help fellow robot hobbyists!

A. This is a possible answer for Roger's problem getting a precise position for his robot. I have also been looking for a system for precise navigation for my outdoor rover robot.

Combining visual odometry and landmark recognition could be a solution. Check out www.cs.cmu.edu/~rahuls/ pub/icra2005-rahuls.pdf. It might not get as close as an inch, but I think 10 cm is a reachable goal!

Needed for this approach is a computer with OpenCV software and a webcam. I built myself a Linux machine based on a cheap mini-ITX board. I'm still in an early stage, but it is promising.

—Allan Bertelsen Denmark

Another reader response:

A. I read your December column as usual and have a suggestion for Roger (who wanted a system for locating a robot in a 1/4 mile area).

Your idea to put a light source on the robot and cameras at the property edge is a good idea, but it can be far more effective if you reverse the placement.

Put a camera (or other detector) on the robot and then place any number of beacons (the number depends on the accuracy you need) throughout the intended environment. Each beacon could be coded with different colors (as you suggested) or different modulations of one color (perhaps infrared, depending on your sensor).

Alternatively, every beacon could be the same color or modulation, but give the robot the ability to remotely

14 SERVO 02.2009

turn on and off any beacon of its choice.

Either way, the robot could triangulate itself (as you mentioned) with any two beacons, giving better accuracy. It could also just move toward a certain beacon for a given distance, then turn towards another beacon, etc., letting it navigate along a path it creates (made up of movements toward beacons). This is explained in a book I wrote with my co-author (Robot Programmer's Bonanza, Chapter 15) where we used a series of beacons to control navigation in a suite of offices. The robot maps out a path (using algorithms similar to those in automobile GPS systems) and then turns on the required beacons one at a time, to move to the desired destination.

Probably the most powerful aspect of this algorithm is that it can be used with any type of beacon (visible light, infrared, RF, etc.) and then can be adapted to a large number of beacons in order to cover huge areas or to increase accuracy.

We wish someone would produce this product line (a sensor plus any number of beacons); the code for each beacon could be set perhaps with an eight bit DIP switch. In its simplest form, the beacons would be just bright colored light that could be controlled with an RF link from the robot. A standard webcam could be used to detect the light (just turn it off to verify proper detection). With a beacon hung at the top of each doorway (and other appropriate places of interest), navigation of a house would be easy.

RobotBASIC — the language used in the book mentioned above — provides the ability to access all ports (for controlling the beacons), commands to capture images from a webcam, and even commands for locating specific colors etc., in a user-defined grid pattern superimposed on the image.