Anyone ever do the math to figure out the absolute minimum amount of energy (per payload mass) necessary to balance in 2 or 3 dimensions?



Anyway… this Honda unicycle segway thingy (the "UX3…watch the video!") uses those bi-directional wheels (that would turn Occam into the Barber of Seville) made of sideways spinning roller skate wheels mounted around the circumference of a larger wheel.



All of that is secondary to the fact that Honda has here repurposed the balancing algorithms it originally developed for its ASIMO humanoid robot.

Notice that the design ergonomics of this thing have the rider's toes skimming the floor. I suspect this is to allow turning (drag the right forefoot to swivel to the right). These bi-rotational wheels can't effect a swivel (orientation change) on their own. On a regular unicycle, the rider effects a turn by twisting his body rapidly to overcome the friction between the tire and the floor that would otherwise tend to keep the wheel's rotational direction constant. If one had separate control the rotational rate of each of the 50 or so skateboard wheels at the wheel's circumference one could effect a swivel (but in this case, Honda seems to be turning all wheels in concert). If they did have variable speed control of each wheel, they could collect torsion measurements (between the foot pegs and the seat) to initiate rotational commands.

I am assuming that Honda is using a set of counterrotating flywheels which would dampen balancing jitters. The same flywheels would work against rotational turning schemes of any kind.

Of interest here is the fact that Honda's balancing algorithms must be crunching some (particularly nasty) version of the "double pendulum problem". I wonder how much Honda's algorithms lean upon the rider's own balancing instincts (what would happen if the rider lifted his feet off of the foot pegs?).

I did a google search for "self-learning balancing algorithms" and found nothing.

Oh, and, yes! I am embarrassed that I have wasted my time (and yours) on meme candy like this when there are some slightly more interesting and more challenging obstacles standing between today and real general AI than what suffices as its body or what wheels it rides to town on.

Side note: What crazy motion feedback loops are to be expected when you mount a mass shifting actuator on top of a balancing platform and have to control both with the same computer chip?

Randall

Cool. But all I see is something that won't get me there faster than I can bicycle. I guess I'm more impatient than lazy.

I certainly didn't post this because I am interested in the technology. I posted it to expose the lack of direction in high tech.

Randall

Cool. But all I see is something that won't get me there faster than I can bicycle. I guess I'm more impatient than lazy.

I certainly didn't post this because I am interested in the technology. I posted it to expose the lack of direction in high tech.

I posted it to show that awkwardness and hard causal limitations exist (surprise) in the world of the physical, just as they do in software and logic.

Randall