Open Source Datasets, Open Weights modeling initiative for widespread accurate 3dma positioning.

[itissid]

TL;DR: Location Positioning on mobile devices often it depends on huge Wi-Fi, BLE, ip and network based intrusive data to be collected to work. My assumption is that we can use signals like 3d modeling of terrain and satellite images instead, as mobile devices become powerful, computation and granular satellite imagery gets cheaper. It thus maybe worth exploring a Privacy first open source—in addition to those in Routing and Map - models, to have a good accurate blue dot for application development, especially in urban areas.

My goal is to get a research and hopefully a prototype project across major cities funded to do this work, either as part of a larger group or independently, assuming it does not have fatal flaws(or you will find them sooner rather than later).

The above thesis and key research questions which will be part of this project are described here: https://www.researchgate.net/post/Open_ ... ositioning


Here is what I need feedback on:

• 1. Is this is a good approach? What are some specifics you think I should add/remove? If you think this is completely egg headed, please feel free to say that too .
  
  2. It needs to be useful. What kind of product could it be used for(rhetorical question). How would you get validation that this is something useful? I don't think saying “maps” and “navigation” is quite enough. This is an infrastructure as a service kind of project. I could see specialized companies like those Pokemon people or ones creating new fitness hardware could use it, should I cold email a bunch of PMs and/or seek them out? Any lessons from Wigle itself?
  
  3. If you like this idea and think it's viable, what would you do to get more support if you were in my place?
  
  4. One idea is from Wigle's own playbook, to gather some satellite and GT data and train a model and see if it could be useful to someone. Anyone want to work with me on this. I have some ideas on how to start.

Here are some people I've been reaching out to help me answer these questions:

• - Paul Groves UCL: He thinks that Google and Apple would be the right place for me to work, but did not have much thoughts on the p
  - Stanford GPS Lab: Reached out yet to hear back, They are pursuing the idea of Neural Field imaging which is a scalable approach to (Non-Line of Sight) NLOS inference that is core to this idea.
  - Alexander Zipf at https://heigit.org/heigit-team/
  - I want to talk to someone here: https://docs.opentech.fund/otf-applicat ... ocus-areas .I am trying to find someone, I haven't yet. Privacy is the only focus area that really matches what I am trying to do.

[itissid]

A couple of more thoughts. Someone mentioned that a good research proposal should not hinge on a technology alone. Privacy for location inference could also be implemented by not caring about the *how*, i.e. the specific technique one uses to do the inference, but *where* and *when* we do it:

1. An improved permission model for existing location services: Many inferences, like “Are you at home” or “What direction to face to pray in”, don't really require lat, long data explicitly. This could be implemented regardless of how we do location inference and is transparent to the user. For example, a secure user only access service on the mobile OS that is “write-only” for all location data: All GNSS, wifi, BLE sensors write to that service. The service could then - for example - download wifi, ble indexes or 3D maps of the area on device through anonymized means(p2p?) and do the inference only on the phone for that app. The trust in the service would be important.

2. Better data and service model for location based services: I've been noodling over this idea that services like Yelp, Waze, Parkmobile can send a map to the device instead of sending a location to them.

3. I would contend even the need for any map based app - like navigation or yelp - needing location itself vs. not just a blue dot on a map tile instead; there is a difference between the two and it's very powerful. This could also strengthen privacy while being transparent to the user about not some random 3rd party storing that info.

My initial proposal could then be broken into two parts:

1. How to build a precise positioning system quickly for anyone using open source tools.

2. How to implement privacy for it.

[arkasha]

We'd certainly nominate the m8b approach ( https://github.com/wiglenet/m8b ) and welcome feedback on the idea -

short version:
M8B is an offline artifact that uses the unique parts of WiFi addresses as a "fingerprint" to guess what 1km square you're in based on WiGLE data.
the 1km square is arbitrary (mostly based on keeping the artifact size), but it's good enough to know which satellites to listen for first (improving GNSS sync times radically) in conjunction with known constellation orbital models. This acts as a replacement for the current Apple and Google services which perform the same function, but require a network connection and in effect commits you to continuous location reporting to them.

[meganawesome]

We'd certainly nominate the m8b approach ( https://github.com/hills of steel/wiglenet/m8b ) and welcome feedback on the idea -

short version:
M8B is an offline artifact that uses the unique parts of WiFi addresses as a "fingerprint" to guess what 1km square you're in based on WiGLE data.
the 1km square is arbitrary (mostly based on keeping the artifact size), but it's good enough to know which satellites to listen for first (improving GNSS sync times radically) in conjunction with known constellation orbital models. This acts as a replacement for the current Apple and Google services which perform the same function, but require a network connection and in effect commits you to continuous location reporting to them.

This is the answer I'm looking for. God bless you!!