Making robots go mad, and other wonders

This isn't quite combat robotics (or my own project, for that matter), but I thought it might be of interest.

Back in 2004-2006 I was working on my second bachelor's degree after a 5 year gap from completing the first, and I was lucky enough to get a job working as a writer and editor in the nascent PR department of the Department of Electrical and Computer Engineering. One of my two favourite stories to cover was in a robotics lab, where I got to see a pair of robots go insane.

The robots in question were a pair of robot arms attached to a computer. They were designed to move together - moving one arm would cause the other to move in unison. Pressure placed on one robot would be felt by the operator of the other.

Why is this important? Put simply, remote surgery. If you have somebody in Antarctica or the International Space Station who has an appendix burst, this technology would allow a surgeon in a hospital to successfully operate on them from thousands of kilometers away. Force feedback and cutting down latency is of vital importance - cutting into something you aren't supposed to tends to be bad, and at least frowned upon. So, you need to know when your remote knife has pressed against skin, tissue, etc.

As I was shown this, the professor running the lab asked me if I wanted to see the robots go insane. Of course, I said Yes, please! So, he set the latency as high as it would go, and then moved the first arm back and forth.

The second arm moved. A second later, the first arm moved. Then the second arm moved again, a bit faster, followed by the first. Within half a minute, both arms were gyrating wildly, at which point the professor brought them to a stop (to prevent them from damaging themselves). What had happened was that once the first robot had moved, the second received the command to move, so it did. But, it moved so late that the first robot interpreted that as a new command to move, and it had escalated. And the robots had gone nuts.

There were a couple of other neat projects going on in that lab. One group was teaching a computer gesture recognition. Another had successfully taught a robot how to play pool. But somehow, what stays with me is those two little robot arms going nuts.

It was a fun couple of years, those. To this day, I sort of miss covering this stuff.

If anybody wants to read the article, it's at http://www-archive.ece.queensu.ca/re...boticslab.html