The goal would be to accomplish the exact same thing that camera tracking does, but it would be less time consuming and you’d never have bad tracks, even if your scene has no good track points or moves too much.
I haven’t seen anyone do or talk about this, but I can’t imagine that it’d be any harder than motion capture (I also understand motion capture accomplishes a very different goal than this, but I think that the process would be similar). In fact, it should be easier. Instead of having so many points moving in relation to each other, you’d just need the motion of the camera in relation to a grounding point. Duct tape an arduino or something onto the back of your phone and start filming.
But I don’t know anything about motion capture and I’ve never done it. With something this simple, I’m wondering if I could just buy a piece of hardware and download a Blender plugin. But can someone who knows more about this point me to exactly what hardware and plugin I’d need and just some basics about what my process might be?