Resurrecting ideas from a 14-year-old journal: the Holonomic Drive

When I was about 10, I realized that I should keep a journal of my ideas.

Mostly, I started doing this because other people were having the same ideas. Many of the things I was sketching out in class were appearing in Popular Science and the like a few months later, so it seemed I was on to something(s).

Many of the ideas were robotics-related. I'd been tinkering on mobile robotics off and on since I was about 8, thanks to my techie parents.

Around age 11, I started to think about alternative locomotive schemes. Most robots I had encountered were driven by wheels, obviously inspired by a car or wheelchair. As anyone who's driven a car knows, the wheel scheme limits your mobility (three-point turns, anyone?). This is even more painfully obvious to anyone bound to a wheelchair.

I had seem walking robots. Hell, I had built walking robots; the robotic spider I built out of Legos in sixth grade was slow and fragile, but it freaked the hell out of the girls. I learned a lot from it — for example, I learned that building creepy robots is a great way to put off getting laid until college, if not later. But more importantly, I learned that man-made legs are complex, error-prone, and fragile. (For the other armchair roboticists, I used a pantograph leg mechanism with overlapping sweep and tripod gait, each with two degrees of freedom. Not that I knew any of those words.)

So, during summer school, I came up with a number of alternative schemes. One (which I called the Trinity Motor) involved special wheels which had smaller wheels, pointing off to the sides, around the rim. Here's a cleaned-up version of my original sketch:


Envision the axle coming out of your monitor, through that hole in the center. The whole assembly can rotate clockwise/counterclockwise (rolling to the left or right). However, if you grab the axle and push it into your screen, or pull it out of your screen, the little wheels on the edges roll freely. The key here is that the wheel can push along one axis (left/right, here) but be pushed freely on another (in/out of your monitor).

So, then you gang three of these up, 120 degrees apart around a circle, and you get this:

In this diagram, two wheels are being turned by motors (in the directions of the colored arrow) while the third wheel is being pulled along freely on its little edge-wheels.

This type of drive can move freely in any direction (well, except into the air), and can even rotate while it moves: if each wheel in the diagram spun a little more counter-clockwise (or a little less clockwise), the whole body would turn as it moved.

This was before the age of cheap and ubiquitous computing power, so I even designed a special joystick to control the base, doing the calculations with a series of interlinked levers. (I now understand that the calculation is the "vector dot-product.") The main problem I could see is the bumpy ride: as the wheel turns, it would go thunk thunk thunk from edge-wheel to edge-wheel. This can be reduced by putting more, smaller wheels on the edge, but that struck me as less-than-ideal.

Anyway. Shortly thereafter, I discovered girls, and my productivity tanked. Fortunately, however, someone in the past n years has had the same idea, and fixed the wheel problem!


Clever! I exclaimed. Two of my wheels, bonded back to back, so that a roller is always touching the ground!

In the mainstream robotics community, these are called omni-wheels, and the drivetrain design is known as a holonomic drive — a fancy math word that means it can rotate in place, move in any direction, or do both at the same time.

My original prototype used scavenged rollerblade wheels on wire around a cut-up lawnmower wheel. These prebuilt wheels are not only higher quality, but at $16, they're cheaper than building it myself. I think it's time I return to robotics.