nRF51822 beacon kit: big notches in RSSI vs distance plot

[classicman]

Hello everybody,

I'm designing an app to track the position of the smartphone using six beacons and, in order to do that, I made some measures both indoors and outdoors and plotted a graph of RSSI vs distance using the data acquired from nRF Connect app (exported to .csv and fed to Matlab). The first set of measures was taken in a corridor about 3m wide and 35m long at different transmission power level, the second set was taken outdoors in an open field about 30x25 m at 0 dBm transmission power. Both the plots show serious notches, as you can see from the pictures, causing great troubles with the accuracy of the tracking.

I'd like to know if this behaviour is to be expected or there is something wrong in the setup of the measures.

Thanks.

[classicman]

Thank you both for your replies.

While the data is not raw since the nRF Connect app itself plots the average value of the RSSI over different packets received, I do agree that more sophisticated filtering along with additional information might be of great help to calculate the position of the smartphone.

I guess I still need to figure out how to proceed.

But first, since I was asked for a notch-free plot, I'd like to remove those. Do you think the graph could benefit from taking measures in an open field (like, an actual field, completely free of obstacles in every direction) or the reflections of the ground will always be a problem, and I should focus my efforts towards the filtering only?

Thank you all again.

[Sigurd]

You can find a lot of resources on this topic online. A common approach seems to be to apply a Kalman filter.

wouterbulten.nl/.../

www.youtube.com/watch

[endnode]

I believe this is domain of Location Service (or RTLS/PLS) solutions which are already explored and offered by several companies for several years. I'm afraid you can spend years of testing and researching in the lab to arrive to pretty much equal conclusions as all these people... also note that there are several patents in this field so whatever smart you want to use you are in danger of IP litigation (if you want to sell it as product or service on the market). Sorry that I'm not helping much but this is conclusion of our 6-12 month side-project in this field (we did it in 2015 already).

[Roger Clark]

Basically RSSI is never going to give you an accurate distance.

If you have lots of beacons and know their exact topology, you could mathematically process the RSSI data, give you a more accurate position.

However I think that the generic maths behind such an algorithm is probably something to ask on another forum.

BTW, Like @endnote, I have also investigated indoor positioning and came to the same conclusions. i.e Using RSSI will not give you an accurate position. I also investigated the use of accelerometers and gyros and magnetometer data to augment the RSSI but it introduces another set of problems.

[classicman]

Thank you all for your insight (and useful links).

The app would not be sold, it's just a University project, so at least that is not an issue.

Yes, I do have more than one beacon available, in fact I was planning to use six for a single room.

May I ask @RogerClark, what problems did you meet when trying to interface with accelorometers? Because that's something my professor mentioned, using accelerometers in aid of the RSSI.

Thanks.