A step, indeed; this is research, it need not be actually useful. The things you mention, I can see as "left as an exercise for the reader".
I am more worried about power efficiency. Corrections welcome, but I think accelerating and (rapidly, in many cases) decelerating a flywheel can't be that efficient. Also, I don't see how those flywheels could efficiently move a large assembly of these things. Let's say you have a 5x5x5 cube: how would they work together to move that? Fall apart and reassemble?
I actually think that spinning up a momentum wheel is fairly low energy, but it's been a long time since we did that.
Like I said, if they locked together you could get movement from a huge cube. But I think the key in these guys is to build dynamic rigid bodies and have "joints" that move around to connect them and help them actuate. That'd be awesome.
They could also snake along the ground doing a chain where one goes from the back to the front, then the next and so on. Anyway, pretty cool idea to start from. I can't wait to see if they iterate this and add cameras and a main controller or anything like that!
I am more worried about power efficiency. Corrections welcome, but I think accelerating and (rapidly, in many cases) decelerating a flywheel can't be that efficient. Also, I don't see how those flywheels could efficiently move a large assembly of these things. Let's say you have a 5x5x5 cube: how would they work together to move that? Fall apart and reassemble?