There are some cool research papers about delayed processing of GPS data in the cloud. The idea is you turn the GPS on for just a few milliseconds, record the raw radio data (without getting a GPS location fix), and do that every 10 seconds or so.
Then later you upload all the collected data to a big cloud compute cluster which can figure out all the locations (and where battery life doesn't matter).
People are using that technique to have GPS trackers with years of battery life - handy for things like tracking animals.
Tracking animals would seem to be a different set of requirements than, say, turn-by-turn car navigation.
I'd imagine that for a lot of research, longer lifetime would win over real-time ish data, and possible you don't care so much about precision and granularity either. You probably want to upload semi often or risk losing the whole thing, but otherwise minimize batter use.
GPS already isn't good enough for car navigation, which also integrates wheel rotation and steering angle sensors. Even on a bicycle the tracking is noticeably better when you add a wheel rotation sensor to a GPS head unit.
Pretty high - each location sample is hundreds of kilobytes if I remember correctly, although it was possible to trim that down if you knew there was a strong signal or you were happy to have some probability of an incorrect location.
There are some cool research papers about delayed processing of GPS data in the cloud. The idea is you turn the GPS on for just a few milliseconds, record the raw radio data (without getting a GPS location fix), and do that every 10 seconds or so.
Then later you upload all the collected data to a big cloud compute cluster which can figure out all the locations (and where battery life doesn't matter).
People are using that technique to have GPS trackers with years of battery life - handy for things like tracking animals.