I went to a meetup with my son Kris last Sunday. The topic was vision for autonomous vehicles.
The topics were very interesting, and mentioned some techniques and equipment that I might be able to use in the future (Lidar Lite). Of course, the best part is meeting new people interested in the same things as me.
One thing I have given quite a bit of thought to - using the 4G cellular network as a control/fpv system. It should be possible, with Verizon among others, claiming sub 200mSec latency. Apple FaceTime is almost good enough to use as a fpv (First Person Video) system just as it is - but seeing what the 'copter sees doesn't do me a lot of good if I can't control it.
Using 4G would be so much better than trying to use ever-increasing transmitter power to send video signals back to the "base station". Theoretically, the distance would be nearly unlimited. Yes, most of my 5 aircraft have GPS, and the ability to fly autonomously, but it is better if you can see something from the air and then go over to explore it more closely. It is also handy if an obstacle is in the (pre-set) flight path. If you can see it, you can probably avoid it.
Using 4G as a control method also would solve the problem of controlling the craft at long distances. I could theoretically fly a quad around Chicago and capture the pictures (assuming some friend was in Chicago to handle the battery charging).
I suppose I could develop a modulation scheme that would take the output from the joysticks and encode them on an audio stream, and then a decoder at the other end to split out that audio signal into individual channels which would control the aircraft. That way I could use Face Time pretty much as-is. I would get my signal from the audio jack of an iPhone that was carried by my aircraft and would use the iPhone's camera for the vision, but I really worry about reliability - I have had FaceTime videos freeze. If the control system also froze, that would not be so good.
Maybe that is when the "autonomous" features could take over for awhile and do some fairly safe maneuver, like hover. If the signal disappeared for too long, it could just return to the takeoff point and if it got partway home and the signal was re-established, it could be under manual control again.
I guess that I would really like to buy (or help develop) an ARM-based device (or one device for each "side") that had a RTOS (Real Time Operating System) that was custom designed to do only the tasks I outlined, and to do them very well. I think there are some real possibilities, and it shouldn't be too long before someone comes up with a good solution.