Sensor fusion is not new. In fact if you look at what amounts to sensor fusion in this case, it's still heavily based on satellite timing, not a terrestrial or celestial reference. RF transmissions don't (generally) come with directional data embedded so just having those as locus doesn't help, you need to be able to third reference.
For example you can estimate your location to a cellular BTS based on CRO, however that only gives you distance not azimuth, it's only if you have three towers that you get accurate localization - and even then from the signal alone you don't have a global reference (ie Lat Long or MGRS).
The major overriding issue with geolocation, or localization, is having a consistent reference. Celestial reference is still the best bar none reference. GPS is the best DIGITAL reference because of it's simplicity and accessibility - even though it's accuracy is pretty poor really. Until someone comes up with a more robust, or more easily digitizable consistent reference, then everything will be based on satnav.
Nope. In fact, this is the oldest method of radionavigation available. A beacon of some sort sends a radio signal, and a directional antenna identifies the bearing to that signal.
Toss in some sort of embedded signal (like, say, a tone-based code which uniquely identifies a beacon), add another such beacon, and you can fairly reliably locate yourself in a two dimensional space.
Throw in some form of DME transceiver (it receives a radio pulse and returns it with a fixed latency between) to give you a distance to that tower, and you don't even need two of them; just a compass.
I honestly wonder how long until some enterprising soul starts implementing VOR and DME at a city block level...
I'm missing the significance of this. Don't many devices (especially cell phones) already do this with Wifi & cell tower info? Is the innovation here that there is no GPS receiver used at all? How will that work in rural areas with less signals to use?
No, this technique is different. What cell-phones do is rely on pre-mapped databases of wifi and cell tower locations to do trilateration to get a position.
The technique the article describes is basically using any stationary RF source as a reference point to tell if your position estimate is drifting. (It's not unlike using a mountain as a landmark for your direction and location while hiking; you don't have to know precisely where the mountain is, you just have to be able to see it and know that it's not moving.) As you point out, it's dependent on what RF sources are nearby (this is the "opportunistic" part of their SOP signal acronym) but can be used to help compensate for the drift that inertial navigation systems have without having to have a pre-compiled database of RF transmitter locations.
Wifi and cell towers are used to transmit known positions to the device. This observes and tracks the positions of the signals in real time and then determines the changing position of the device from that info.
I skimmed some of the paper, they discuss it in terms of maintaining accuracy of inertial navigation during loss of GPS signal, not in terms of never using the GPS signals.
The cell towers transmitting location data is how assisted GPS already works, the wifi part is how Google does fused location services on Android (with all the other sensors).
What improvements does this offer over the already sensor fused (GPS, Wifi, Cell, compass, barometer and inertial sensors) location services Android provides?
Add this to another reason why traditional GPS "upgrades" for cars are no longer worthwhile. The other reason being that smartphone navigation works just as well if not better than most car companies' custom systems.
It is nice to have if you're an American driving in a different country, though. Maybe not enough to justify the cost, but very helpful to have. American phones are basically crippled within 100ft of the Canadian border.
As someone who works, studies, and has published in this area or work, there is nothing new or novel about what they did. They only repeated work and ideas that have been around for a while and hyped it up.
For example you can estimate your location to a cellular BTS based on CRO, however that only gives you distance not azimuth, it's only if you have three towers that you get accurate localization - and even then from the signal alone you don't have a global reference (ie Lat Long or MGRS).
The major overriding issue with geolocation, or localization, is having a consistent reference. Celestial reference is still the best bar none reference. GPS is the best DIGITAL reference because of it's simplicity and accessibility - even though it's accuracy is pretty poor really. Until someone comes up with a more robust, or more easily digitizable consistent reference, then everything will be based on satnav.