Manduca DARPA Project

Flying is a difficult problem in an integrated system. The question is at what level of neurobiology could our course/coarse control (100 bits, 10 times a second) steer a moth? Will it be lower on a descending neuromuscular nerve bundle, causing the moth to jerk to the side and fall five feet, but regain a course that is say 30 degrees to the left, with a standard deviation of 30 degrees, so that it barely staggers and zigzags to the target, or can we hit the brain at the right spot to make it fly at exactly the azimuth we tune our dial to, or somewhere in between, say the cervical connective, so that we can bias the flight in some kind of nonlinear function --which might require a skilled operator not to over-correct in this situation, like sailing a boat?