Mapping Interior Spaces With Robots And GIS

Roland Piquepaille writes "In an article about GIS and Robotics, Directions Magazine reports that architects and other professionals can now use spatially intelligent robots to collect interior space data. With such mapping robots, it's possible to capture accurate data for over 10,000 square meters per day and to easily integrate it with existing software. The article doesn't mention the sources for its illustrations about these robotic systems, so I thought I'd point them out: a company in Maine called Penobscot Bay Media. You'll find more details and pictures about these mapping robots at ZDNet."

Comments

Re:TFA: Don't be a robot, use one!

[auric_dude]

The Air Force Research Laboratory may well be already using robots to map out underground enclosures http://www.wired.com/news/technology/software/0,71 779-0.html some more information and ideas discussed by David Hambling via http://www.defensetech.org/archives/002775.html

Re:spam spam spamity spam

[onion2k]

Your 'fuckroland' comment + the pennyarcade image you have linked as your homepage = Amusingly ironic.

A shameless plug :)

[Soul-Burn666]

As a small semestrial academic project, I worked on a different kind of mapping project which uses a large number of very simple (and cheap) robots instead of a small number of expensive robots like in this article.

Each robot is aware of its location through odometry (measuring the distance traveled by both the of the bot's wheels) and collision detection using, in our case, a rotating straw due to the fact we were limited to Lego Mindstorms.
Using odometry inserts a lot of error to the calculations. To counter these errors, the robots communicate over a short distance (touching distance) and average their expected location and heading.

In theory, and simulation, the algorithm proved very successful. Especially for a large number of agents.
In practice the errors were too large compared to the very small number of agents (4) we had at our disposal.

The project page.
And the simulation applet, written with NetLogo.

I wonder if they use such averaging algorithms with these robots aswell.

Re:A shameless plug :)

[w33v1l]

One technique for dealing with errors and gradual drift is the use of a covariance matrix to allow the propagation of location information through a scene. Such as in Andrew Davison's work http://www.doc.ic.ac.uk/~ajd/ , although this only deals with a very sparse map of feature points, using mono vision in real-time. One nice thing about his SceneLib and MonoSLAMGlow software, is that it'll work with a relatively cheap single webcam.

Another nice visual SLAM (Simultaneous Localisation And Mapping) page is Robert Sim's http://www.cs.ubc.ca/~simra/index.html , where he uses multi-camera vision to produce a much denser map.

Alternative: Time Difference of Arrival

[Ceriel Nosforit]

I've been thinking about implementing Time Difference of Arrival plus echo-locationing to do a very similar thing. Plus, this method could be used to scan in 3D with ease. Not to mention a number of other interesting applications, such as beamforming plus distance awareness.

The very same code could be used for EM for a sort of total situation awareness radar, passive and active.

Next up: death ray! Mwahahahaah!