Being an engineer with lots of control system experience and a T1 for 59 years, I've been thinking about this for many years.
There are definately issues to overcome. One is the accuracy of CGM systems. Another is the lag between true BG and the reading on the CGM. Another problem is that the way the body reacts to insulin constantly varies. If it was always linear and we have perfect BG sensors, the artificial pancreas would be rather easy to develop. I think that to truly work well the system needs to take into account much more than just BG. Probably things like heart rate which would could give much more information about a person's current state.Rate of change and lead/lag times for transient events also vary all over the place. I also think that the best system wouldn't use a digital computer to do the calculation, but rather a neural computer (one that learns like the human brain) and can learn the way a body reacts to many stimuli. It's pretty obvious that a closed loop system requires the ability to both raise and lower BG. Years ago when I was working on a simulated design for an artificial pancreas we would take many BG readings, toss the high and lows and average the rest. I believe that just like current CGM the system will have to be calibrated with finger sticks or some other method.
The idea of incorporating it into a Smart Phone while obvious presents a whole host of challenges, To the best of my knowledge the FDA has never allowed such integration A few years back I had a meeting with the CEO of Insulet to discuss a product where an add on to an iPhone could serve as the PDM for and OmniPod. It could definitely be done, but the FDA at that time had some stiff prohibitions against using the same power source that served a life support role also powering other unrelated things, There are also the issues with a problem, be it either a virus or bug in the OS or other application causing a problem with the life support software's operation. (I did come up with a way to work around these requirements) That being said when they are available I'll be first line.
I discussed this subject with my Endo this past week. She says a truly closed loop will never be approved in the USA. The best we can expect is low & high blood sugar alarms. She also indicated that she had spoken to somebody from omnipod recently and was told that the high/low alarms incorporated thru a CGM with Omnipod is something that is going to happen very soon. Who knows!