If you look at the design, you'll notice that there is a defined 'front, side and rear.' A pair of engines/blades control up/down on the front/back, basically acting as an elevator for climb/descent with most of their energy used for lift. Higher rpm from the rear of the pair and the nose points downward a little. But all six are being used for maintaining/changing altitude or increasing 'forward' speed. Higher rpm on the front one and the nose points up as needed, while all six are still used for maintaining/changing altitude or decreasing forward speed. The other four engine/blades act mainly as the lift source and marginally as roll control/changing direction of flight.
This setup wastes very little energy since all six engine/blade combos are almost always mainly working to lift. The second huge advantage of this arrangement is that the vehicle acts more like a 'standard' aircraft than a helicopter. Note that the blades rotate in different directions, that negates the problem of rotation around the vertical axis that causes problems for most helicopters. Secondly, with a definite 'front, side and rear,' the GPS data can be used as a simple compass to keep the vehicle pointed in one direction or another through the built-in stabilization system. That stabilization system would then need only the two accelerometers for holding the desired attitude. I think this setup would be simpler than using three accelerometers for movement and the GPS simply for returning to a given point. And probably cheaper, also!
But I have no direct knowledge.
Whatever design was used, a
big problem is fine tuning the the sensor input and output with just the right amount of dampening to prevent self-amplifying oscillations. In a real aircraft, this can happen when the pilot gets slightly behind the aircrafts reactions to control inputs. It is usually seen in the pitch axis with the nose becoming ever higher and lower as the pilot inputs lag the actual aircraft position. A good pilot can recognize this and correct it before it becomes dangerous. A great test pilot can get the aircraft back on the ground safely so he can "discuss" the problem with the engineers!
But with an autonomous vehicle, all this recognition, control movement and dampening must be done by the computer(s), assuming the vehicle is to be allowed to actually maintain a fixed attitude/position/movement for any extended time. My biggest problem with all the RC aircraft I ever flew (or attempted to fly!) was recognizing the reactions of the aircraft purely visually, from outside and often at a considerable distance. The control box (transmitter) usually had absolutely no feedback except for the, "You can't push that lever any further, stupid!" kind!
And, of course, one got even less sensory and visual feedback from what one would get in the vehicle (blood rushing to one's head or feet, one's head slamming against the canopy, etc.).
So, it looks like this vehicle has been very well designed to limit it's range of aerobatics to those of "grandma!" Or as we used to say, "the excitement of flying a B-52!" Straight and level, no barrel rolls, Immelmanns, split-S's, etc. "Down" is almost always completely toward the center of the Earth, up is always the opposite of "down." "The nose/tail/wings shall never point at either of those directions!"
The designers seem to have included just enough stabilization to keep the vehicle very stable, vertically so that camera work can be done easily. I doubt they considered carrying liter bottles full of liquid as an appropriate use! But it seems to still have excellent stability, even with a moderately free swinging load
Of course, that still allows for fairly rapid horizontal movement and the main purpose of this machine; vertical movement!
I'd vote for this thing as extremely well engineered and very clever! Not cheap, of course, but neither are MQ-1s or 9s...
