It's computationally expensive, basically. If you're in to that sort of thing, check out plasma pong (I think you can google it). Maintaining and updating velocity vectors in a huge vector field is really cool, but it would simply bring the program to a crawl.
What if it were very ineficient and inacurate?
I haven't worked it before, so I'm going on heaer say. If the user is running some multicore computer, it might be easier. This sort of thing is pretty easy to parallelize.
Say 10 away on the eye return, and only with 1 degree of area. THis would mean a jet could be used. The physics for emulation shouldn't be too difficult, since only a small area of vectors would need to be created; we can also assume that the jet affect carries no true inertia, so it would stop once a bot turned.
I don't understand exactly what you mean here. Are you suggesting that we
just have lines of force from a bot's motion, ignoring issues of turblence, etc.? It might work. It would simplify things, certainly. If we simulated a sort of elongated triangle off each side of a moving bot, representing its wake, and an elongated path where the bot has just been, representing the dead zone inside the wakes, and slowly expanded the shapes to represent wave propogation, it might work.
So I dunno, maybe.