Hacker News Reader

Comment by modeless | Hacker News Reader

The Kinect-mapping quadrotor was the most interesting demo; that could be useful today. Imagine letting loose a swarm of quadrotors in a large building and having a complete map and textured 3D model of the interior and exterior within just a few hours.

View on HN

The Kinect has a very limited range, is not accurate and works only in certain lightning condition. That's why they added a laser scanner. Those are over $1000. Then in order to build the map via SLAM in real time you need a really powerfull PC/Laptop. Even a netbook is not going to do it, at least for the SLAM in ROS(ros.org). Then it looks like they added a all angles are 90° assumtion, because normally the map does not look that nice because little errors accumulate. So it will take a while till we see something like that outside universities.

Laser scanners have become a bit cheaper (and smaller) in the past few years (Thanks, DARPA Grand Challenge & Urban Challenge).

Neato Robotics has developed a planar scanner that costs $30 to build--It's on their robot vacuum. See "A low-cost laser distance sensor", http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=4543666...

Hey quit being a letdown, this really amazing stuff. You should reword it more like:

"The video is showing amazing new thingamabobs that will get less expensive over the next few years, and pretty soon you will all be able to use robots like this in your everyday life".

Kinect-like sensors are going to improve very quickly (http://www.theverge.com/2012/2/29/2834088/samsung-camera-sen...), as are SLAM algorithms using Kinect-like data. High cost isn't necessarily a problem for this application, as one team of copters could map many facilities. Computing power can be mostly centralized at a base station that can be as beefy as you want. Certainly a commercial product is years rather than months away, but I think even this limited demo could already be useful for e.g. mapping the nuclear reactors at Fukushima.

For many (perhaps most) buildings, I imagine a simpler wheeled or tracked vehicle would be just as effective and surely cheaper, more reliable, and more robust.

Why? These robots look pretty simple and have no problem going up stairs.

Navigation for land based robots is a much harder problem. I would expect quadrotors to be more reliable actually (only four moving parts! No ground contact!) with the ability to map areas inaccessible to wheeled robots, and much faster to boot.

It sounds cool, but this seems like a very expensive way to accomplish a fairly mundane and useless task...

The problem with trying to make money using robots is that we still have plenty of people on the planet looking for unskilled work. I don't think we will reach the economies of scale necessary to make robots practical unless that changes.

Um, many industries, auto, electrons assembly have All ready been robotized. People are expensive, have short work cycles and break down often, and require much more maintenance and infrastructure. You have missed the revolution buy a hundred years

"Imagine letting loose a swarm of quadrotors in a large building and having a complete map and textured 3D model of the interior and exterior within just a few hours."

This reminds me of: "I’ve long been interested in Navicubes, which is what I call a small and not-quite-here yet $5 box that always knows where it is, which direction it is facing, and which way is up. Besides vehicle nav, intelligent toys, and robotics, these would open up all sorts of neat tricks, such as a ball you could simply bounce off cave walls for mapping."

http://tinaja.com/glib/funstuff.pdf (Don Lancaster)(The Blatant Opportunist #72)(2002-NOV)

http://en.wikipedia.org/wiki/Don_Lancaster

http://tinaja.com/navcub01.shtml

Reply on news.ycombinator.com