Getting into Robotics: Tobi Delbruck
So, how did you get into robotics?
AxlBot 2002 is the first robot I ever built, in 2002 at the telluride neuromorphic workshop. I knew the only way to learn robotics was to start building robots. I just built the simplest thing I could that no one else had built yet. It was a great help that Mark Tilden was there to encourage this kind of simple development.
The “mother of all disco floors” is the first “production” robot I built. I learned a large part of my engineering skills during this super high-stress development of the luminous tactile floor of Ada, our “Playful Intelligent Space” exhibit for the Swiss Expo of 2003.
Are you still a member of the Magic Castle? How did you get involved in magic?
I wouldn’t really call myself a magician. I just like to play with card techniques. I was first a member of magic castle when I was in high school, and then recently I rejoined the castle so I could go see the performances there, which I enjoy very much. Unlike Davide, I don’t enjoy doing performances unless I am quite drunk! And I never do anything that involves any more than a single deck of cards. But I can do a decent “2nd deal” and “in-the-hands full deck false shuffle”!
What made you branch out into working into robotics?
I would say that the three most important reasons for me becoming a roboticist (although I would hardly call myself that, having never had any courses in robotics) were:
- it was fun
- it has been a great way to demonstrate the performance of our neuromorphic vision sensor, and especially its speed. People like fast robots!
- it’s still fun
What advice for those who would like to consider studying robotics in the future?
So if I would give any advice it would be that you should only do a project that is somehow fun for you, and you should not be afraid of making beginner mistakes. The hardest thing is finishing, but starting is almost as hard. Don’t be afraid to trash what you did wrong and start over with it. One thing I really like about robotics is that you can’t really fake it; if the robot really works it sells itself while it is very easy to spot the weaknesses. And it does help to learn some math; I still regret that I have not yet derived the basic Kalman filter from scratch, and still don’t know what an unscented Kalman filter is!