How to locate(x,y) of a moving device using wifi

[vishal]

Hi
    I am trying to locate (x,y) coordinate of a moving device.There is a central point to which the moving device sends its RSSI value.There is one router between moving device and central point.Since i am getting only one RSSI value from the moving device I cant apply the trilateration method for locating device.Are there any other method for locating(x,y) a device in a wifi network.Please help us

[mleyden]

Worked with this stuff 10/15 years ago: https://www.ekahau.com/

I would suggest you would need to map the area first, noting the signal strength read by the target device, and then use it to plot its approx position. Problem with WiFi is reflections...

Mark.

[Ian.M]

However, with a single access point that isn't going to get you an (x,y) position, at best, it can only give you a range, with a lot of uncertainty.  You can map the signal strength throughout the area of interest but unless you use a ridiculously fine grid (down to 10mm or so) that isn't going to help much with reflected signals off nearby conductive objects that spatially modulate the received signal strength and isn't going to help you at all with received signal strength variations due to receiver antennae orientation or absorbtion by people or objects that have moved since the area was mapped.

Even with multiple access points located at the corners of the area (let's assume a square), each with its own SSID so you can get several RSSI readings, the uncertainty is going to be fairly high.  N.B. the extra access points don't have to be connected to the network - all you are interested in is their position and the RSSI for each.  Turn down their output power so you get the maximum RSSI gradient along the diagonal

[janekm]

Getting distance from RSSI is a fools errand anyway. This visualization illustrates it pretty well: https://m.imgur.com/gallery/jdNA6

Try some optical triangulation like the Vive uses instead.

[Howardlong]

If you have multiple antennas and receivers, you can measure the phase differences of signals using correlation, and from that you can determine direction. If the antennas are widely placed at known positions, and the receivers are time correlated, you can calculate distance reasonably accurately.

Using RSSI is often a waste of time for ranging in the real world of multipath and Rayleigh fading, it's only a broad indicative value. you'd also need to know what the ERP of each source is.

If it's possible to ping a device with an accurately known and deterministic latency, you can range that way too, but I don't believe there's a way of doing that with off the shelf APs. When I've done ranging and velocity measurement (using Doppler) this way, the ping is via a simple RF bent pipe transponder with no signal regeneration.

[tggzzz]

Using RSSI is often a waste of time for ranging in the real world of multipath and Rayleigh fading, it's only a broad indicative value. you'd also need to know what the ERP of each source is.

Certainly with GSM cellphones, the RSSI doesn't give you even a broadly indicative value; it is completely useless! I have 20 year old maps from real-world live systems to prove it Simulations of 60GHz within a room are marginally better, but that's all.

What many people don't seem to realise is that multipath, shadowing and diffraction are required in order for cellphone and wifi systems to work. Sure, the downsides need to be dealt with, but that's the price you pay for basic functionality.

[Kalvin]

Worked with this stuff 10/15 years ago: https://www.ekahau.com/

I would suggest you would need to map the area first, noting the signal strength read by the target device, and then use it to plot its approx position. Problem with WiFi is reflections...

Mark.

Mapping the RF environment beforehand, some extensive probabilistic calculations and using multiple access points may be the way to go. Accurate indoors tracking and positioning is quite hard to do properly.

[tggzzz]

I would suggest you would need to map the area first, noting the signal strength read by the target device, and then use it to plot its approx position. Problem with WiFi is reflections...

... and then do it again when furniture moves, a person enters the room, a door is opened, the curtains are shut...

You get the idea.

[Brumby]

Some things that are missing from the list of requirements is just how accurate you need this to be and in what sort of environment this will be used.  If you did come up with a solution that could locate a target within a typical space (office, home, warehouse) to an accuracy of 10cm, I would think you could retire soon.

Trying to do this from a single reference point is mathematically impossible.  You need a minimum of 3 reference points to define an unambiguous point on a plane surface - and for each additional reference point, you can improve on accuracy and certainty.  Just look at the GPS system.

One fundamental difficulty is that you are trying to utilise a signal that was not designed for the purpose.  I'm not saying it's impossible to do so, just that it's not straightforward and would seem impractical.

[Richard Crowley]

Cell phone tower sites use rubidium clicks to quite precisely measure TIME.  They can measure "time of flight" between the cell phone and all the cell towers that can "hear" it.  So they can calculate a location perhaps within 100m(?), or at least good enough to assign which cell tower to handle that phone.  And, of course the TV-show law-enforcement people can trace cell phone location in real time on their fake PC screens.   

Of course there is no way to do that from the cell phone end. You must be privy to the meta-data of the carrier. And the phone must be within the range of at least three cells.

[tggzzz]

Cell phone tower sites use rubidium clicks to quite precisely measure TIME.  They can measure "time of flight" between the cell phone and all the cell towers that can "hear" it.  So they can calculate a location perhaps within 100m(?), or at least good enough to assign which cell tower to handle that phone.  And, of course the TV-show law-enforcement people can trace cell phone location in real time on their fake PC screens.   

Of course there is no way to do that from the cell phone end. You must be privy to the meta-data of the carrier. And the phone must be within the range of at least three cells.

If you relax the requirement from bog-standard WiFi, then you could always update the WW2-era "Decca Navigator" navigation system. At 2.4GHz you could probably get sub-millimetre resolution

[mikerj]

Could the moving device have a steerable, directional Wifi antenna?

[Electro Fan]

x,y, and z (using UWB rather than Wifi)



https://www.zebra.com/us/en/solutions/location-solutions/enabling-technologies/dart-uwb.html

https://ubisense.net/en

[SeanB]

Cell phone tower sites use rubidium clicks to quite precisely measure TIME.  They can measure "time of flight" between the cell phone and all the cell towers that can "hear" it.  So they can calculate a location perhaps within 100m(?), or at least good enough to assign which cell tower to handle that phone.  And, of course the TV-show law-enforcement people can trace cell phone location in real time on their fake PC screens.   

Of course there is no way to do that from the cell phone end. You must be privy to the meta-data of the carrier. And the phone must be within the range of at least three cells.

GSM uses a very precise delay to allocate the transmit slot as well, so the tower can get the distance just by using the difference in the received window period for that particular phone, and use that time into the window as a distance to the phone. Plus using a sector antenna gives a rough area location like DECCA, needing only another tower with known separation to get a transmit block from the phone as well, which gives you 2 distance and arc sections, which typically will only have a small area of error in the 2 distance lines, the arc giving a solution to which overlap it is in most cases so avoiding the lane problem that the omni beacons in Decca gave.

[tggzzz]

Cell phone tower sites use rubidium clicks to quite precisely measure TIME.  They can measure "time of flight" between the cell phone and all the cell towers that can "hear" it.  So they can calculate a location perhaps within 100m(?), or at least good enough to assign which cell tower to handle that phone.  And, of course the TV-show law-enforcement people can trace cell phone location in real time on their fake PC screens.   

Of course there is no way to do that from the cell phone end. You must be privy to the meta-data of the carrier. And the phone must be within the range of at least three cells.

GSM uses a very precise delay to allocate the transmit slot as well, so the tower can get the distance just by using the difference in the received window period for that particular phone, and use that time into the window as a distance to the phone. Plus using a sector antenna gives a rough area location like DECCA, needing only another tower with known separation to get a transmit block from the phone as well, which gives you 2 distance and arc sections, which typically will only have a small area of error in the 2 distance lines, the arc giving a solution to which overlap it is in most cases so avoiding the lane problem that the omni beacons in Decca gave.

You are simply wrong; you should do the basic arithmetic to see how wrong. I know it because I was plotting such diagrams on live data from a local GSM network 21 years ago! See below for the first example which came to hand, shown on a 1:250000 map. The area shown is very roughly 15km by 10km. As you can see, there is considerable uncertainty in location!

The range resolution of a timeslot is ~500m, and the actual timeslot is often +-1 timeslot away from the ideal. In reality, with a significant number of triangulations there is no overlap at all!

Add to that the real-life consideration that the central network operations staff know they don't know exactly what is out there in the field, e.g. they know the cell-site location, but they won't have up-to-date accurate information on all the antennas, nor their sectorisation, nor tilt.

[mmagin]

My recollection is that the RSSI-based WiFi location in my former employer's product line was good for 2-3 meter accuracy in an office environment with optimally deployed WiFi access points.  This is approximately good enough for IT personnel to locate expensive portable equipment misplaced by underpaid employees when they went on break and forgot about it or to locate rogue access points so they could be unplugged.  Latency kind of sucks if you're only polling data from your access points every 5 minutes though.  I think many of the "RTLS" solutions do some kind of streaming feed from the access points.

I think there's some kind of (draft) standard for achieving precise enough time synchronization between access points that you can do time-of-flight ranging, but historically that has been impractical and not supported by the hardware.  Nobody wants to deal with the added cost of say running a seperate piece of coax with a precise 1PPS timing signal to all of their access points -- they just want to use their twisted pair and do ethernet+PoE.

[SL4P]

How about dead-reckoning on-board, using a compass and tach...
Refresh the absolute position only when you know it's accurate.