Thursday, March 22, 2012

Kiva systems

These little robots move the shelves around the warehouse. They claim that it makes the process of collecting the items more efficient.

These are made by kiva systems:

Recently, Amazon has announced that it will acquire Kiva Systems for $775 million in cash. The deal is expected to close in the second quarter of 2012.

Wired has an interesting article about it too. According to this article on Analog Devices' website, the robots use vision to recognize markers on the floor and under the shelves to localize themselves.

Features of the system includes:
  • after a while, shelves with most demanded articles end up on the border of the warehouse, more readily available than less often asked for items. This increases efficiency, and it's an emerging outcome of the system: no need for optimization of the layout.
  • since the robots can work just as well in the dark, only certain parts of the warehouse have to be illuminated, where the human operators are working, thus reducing electricity consumption (but it probably does make up for the energy consumed by the robots...)
  • I like the collaboration between robots and human. It would have been very costly and actually impossible to have humanoid robots go around the warehouse and collect the items on the shelf. Although this would have been more intuitive (more human like), it is a much better idea to move the shelves around. Great out of the box thinking here..!
Altogether this reminds me of MANNA, a fiction novel by Marshall Brain, in which robots end up dominating our world, starting from ... fast food junctions:
"Depending on how you want to think about it, it was funny or inevitable or symbolic that the robotic takeover did not start at MIT, NASA, Microsoft or Ford. It started at a Burger-G restaurant in Cary, NC on May 17. It seemed like such a simple thing at the time, but May 17 marked a pivotal moment in human history."

Friday, March 9, 2012

Working with a webcam

I am currently working on a project to monitor road traffic from cameras. As a result I am playing with OpenCV (and ROS). I bought 2 Altair USB webcams and got problems with them.

Here are some tools to playback video:
- cheese
- xawtv -v 1 -c /dev/video0
- luvcview -d /dev/video0
- mplayer tv:// -tv driver=v4l2:device=/dev/video1

In particular, luvcview has a nice GUI that allows to control video parameters, which turned out to be necessary to turn off the auto exposure mode and adjust it manually.

I had this problem where it was impossible to playback video from both of them at the same time. The solution was to set an option in the uvcvideo driver:

sudo rmmod uvcvideo
sudo modprobe uvcvideo quirks=128
This will be reset every reboot. If this works, create the following file:
sudo vi /etc/modprobe.d/uvcvideo.conf 

containing the line:

options uvcvideo quirks=128

Friday, March 2, 2012


We got a demo from mobileye recently:
A camera mounted on the windshield, detects cars, lanes and pedestrians, and warns you when you are getting to close or drifting out of your lane. The videos were very convincing. We got to test in real life too, although too briefly to really confirm what had seen in the videos.

The product costs about $1000, including installation. It can be installed in almost any car, but this needs to be done by a professional. Some governments give subsidies for it, and some insurance companies reduce your premium if you install one. Apparently some car manufacturers are planning to include the system in their new cars...