Comments (69)
sorted by:
You're viewing a single comment thread. View all comments, or full comment thread.
ATLAS_ONE 4 points ago +4 / -0

Through years of experience integrating GPS into robots.

Buttery 0 points ago +1 / -1

And you can speak to the precision of the data they have on all tracked cellphones at any given time? Don't mean to sound dickish but a lot of armchair experts abound

ATLAS_ONE 1 point ago +1 / -0

I don't understand your question. Are you asking if I have personally audited the kinematic data from every phone? Of course not, I was commenting on the technological capabilities of modern state estimation and how it could be used to obtan a persons location over time in great detail. I don't specifically know the quality of the data True the Vote had access to, but I know what's possible.

Buttery 0 points ago +1 / -1

You demonstrated an understanding of GPS in general and are applying it to data you haven't seen. So I wanted to know if you are certain that this wide-range trawling of location data is that precise. I assumed it was based on pings to cellphone towers, not GPS, as GPS isn't active on every cellphone at all times.

ATLAS_ONE 2 points ago +2 / -0

I'm not addressing GPS alone, I am talking about state estimation in general. Hence why I mentioned RTK and all the other sensors in phones that could be used for state estimation.

I'm saying location can be that precise when you combine various sources of data. The cell tower ping and GPS is just two sources of data. All sensors have error and a level of uncertainty in their measurements. When you combine multiple sources of data together and process them with a Kalman filter you are able to correct these errors and reduce uncertainty in real time. The reliability/accuracy increases in proportion to the amount sensors/data sources you have (up to a point of diminishing returns). If the GPS is turned off then the filter will drop a source of data and location accuracy may be reduced depending on what is left over.

If cell tower pings were the only form of data used then it might be unreliable. I also don't know at what point in the chain that data was taken. When I do this on a robot all the data is collected and processed in real-time on the vehicle. The location data that is spat out from this process is sent back to a control station for viewing. I would imagine a phone would perform the filter on the device itself as well, then send its best guess to whoever it's programmed to (google, network provider, app developers server, etc). Cell tower data alone may bypass this completely by pinging phones on it's network periodically from the network provider end (I don't work with this data so not sure).