MiniPsi

MiniPsi - the MicroPsi robot

MiniPsi is our very own robotic platform. Current robotic experiments take place using Khepera-Robots.

.

They are small and neat, but very expensive, slow and not autonomous. (The price for an on-board camera actually borders on the obscene.) Because of their tethers, it is not easy to have more than one of them in the same environment.

So the natural thing to do is the design of an own platform. We briefly considered the well-priced and sturdy ER1, but found it too large; to have multiple ER1 interact, we'd need to rent a gym. After some fuzzing, we went for designing our own robot, based on a Mac Mini. Why not a gumstix-board or PDA? - Well… we want to run the complete MicroPsi toolkit (including Eclipse and Java5) on it, so we need a complete system with lots of processing power, video input, memory and a large harddisc.

The Mac Mini is compact and affordable, and while it does not come with a battery, it does not consume more power than a laptop (about 12W during idle, 20W during boot-up). So what should it look like?

At first, I thought about using a simple box with the same style and size as the Mac Mini, with internal wheels, so the appearance would be that of Mini whizzing around in unbroken style. But can we fit batteries, electronics, wheels and engines inside such a small compartment?

If we have outside wheels, the thing looks much more "thingish", and we get more space inside. With a larger body, we may also fit an additional display.

Then again, a small box looks so much better!
The next idea was to use an omnidirectional drive. Such a construction is based on wheels that have friction only in their turning direction. With three wheels in a triangular arrangement, the robot becomes very maneuverable, it can move freely in any direction without turning, and it can rotate at the same time.

Thus, the robot will not get stuck in corners, and we do not need to worry about casters as in a two-wheeled design. So the design we settle for looks something like this:

You can see the servos and the distance sensors that will prevent the robot from crashing into things. If we open the box by removing the Mac Mini, the batteries become accessible:

OK, let's build it. Glorious university of Osnabrück, generous institute for Cognitive science, bought us a Mac Mini, which we immediately started mutilating (cutting the power cord etc.). This is the robot with three modified servos glued to a piece of plywood, a controller run by a 6V, 5Ah battery and an iSight camera.

Works. Now for the power input to the Mac Mini. We use 18V, 6Ah and put the resulting 3+1 battery packs on top of the Mac, making it completely autonomous. Look:

We are controlling the servos with a Java API. The camera-image from the iSight can be processed with our toolkit too.

It takes the Mac more than an hour to drain the batteries, and the servos last even longer. Fine. Next we will furnish it with a stylish casing!

Credits

• Kay Berkling and Armin Zundel were there during lunch at the Polytechnical University of Puerto Rico, when the idea was born and then actually made me do it.
• The institute for Cognitive Science at the university of Osnabrück provided funding.
• Nils Appelhans helped in choosing the right servos, tested and modified them and came up with a design for the wheel connection. Jessica Zinn provided moral support.
• Holger Heine from the workshop of the physics department forged the wheel shafts.
• Tim Kietzmann did initial tests with the controller.
• Roland Hafner of the Neuroinformatics group provided experience and a heavy duty battery charger.
• Ronnie Vuine did all the difficult bits, like writing the API, glueing, cutting wires, fitting plugs and putting up with me.