I've considered various concepts for the brake design and think the way forward is to use friction brakes. The original Cubli uses bicycle brake pads, but this has its disadvantages. I'm thinking about using a (possibly bidrectional) band brake for my cube, similar to the one used in the "M-Blocks" [1]. Based on this video and the associated research paper, I'm quite optimistic about this kind of brake.
I suppose there are limits to the size and power of the motors and controllers you want to use, but those inrunners are probably not ideal either, right? I wonder if you could use an oversized hobby outrunner and achieve the same effect (probably with a large controller as well)
Also, even if most hobby motors do not specify Nm/A, they usually give KV, which is actually the same value but shown in a different form :).
I would expect so, but the integrated motor controller might not let you do that. The power MOSFETs and motor windings should be able to handle it. It's all heat dissipation, and this only lasts a fraction of a second.
Those M-Blocks are really clever, the way they move the reaction wheel around to allow it to simulate having three wheels when it has only one. But that limits the 'balancing' capability, though that's - for a cube - mostly a gimmick as far as I can tell, the locomotion capabilities are far more useful and they seem to do that just fine.
There was a popular mechanics article a many years ago about a box that moved without any outward visible means of locomotion. Inside it were a slide with a little tin can and some metal balls, going one way up the slide was easy and nearly friction free, the other way it released and fell, then bumped into a stop at the bottom which was enough to cause the box to slide a bit. That was mighty impressive when I was a kid, this is like an SF version of that.
[1] https://www.youtube.com/watch?v=hI5UDKaWJOo.