So one thing that we wanted to achieve but have not fully tackled it in terms of the visualisation is pulling the closest base station to a beacon. I wanted to create this visualisation as a way of finding proximity rather than exact location.
Based on the experiment found here, drawing the RSSI as a way of determining exactly how far a person is away from a particular base station can be quite challenging as the RSSI values fluctuate too much.
So rather than deciding to create a visualisation that relies upon the distance of a beacon to a base station with the RSSI, like the one below…
…create a visualisation that finds the closest base station to a beacon. Based on that experiment, what it has sort of showed was that it can pick up the closest beacon to a particular base station to some degree. Whilst the results may fluctuate, the further away you are from a base station, the lower the signal gets.
What the basic methodology would look like:
NOTE: Also this methodology is looking at extracting one particular beacon…for now, then I’ll scale up once it works.
- Extract the beacon detections from the database: find all the base stations that has detected the particular beacon.
- Timeframe of detection: As mentioned, the RSSI value fluctuates per detection. So one second, the beacon can be closest to base station 1, but the following second, the beacon could be closest to base station 3 when the person wearing the tracker has not moved. This would mean the person could as well be jumping and bouncing between one basestation to another in one second.
So just as an example, a base station picks up a ibks beacon which has say 2 detections for 5 seconds. As the RSSI fluctuates per detection, suggesting to average the RSSI for those 5 seconds. So rather than having the RSSI fluctuate between -50 to -80 every second, we can average and get a reading of say -60. But how do I know that averaging the values for 5 seconds is sufficient? or 10 seconds? etc Tiara says to do an experiment.
- So after that, extract the highest average RSSI value between all basestations for those 5 seconds - then determine which is the closest base station to a beacon. Then find the second set of 5 second intervals and implement the shortest path algorithm to determine its path.
So another way is tracking someone with areas.
So the image is meant to explain that , the pink dot is where the exact person is, but the person can also be anywhere within the area of the circle. So rather than using location, it also uses proximity to find the approximate area of where a person could potentially be.
However this method also needs the range of which the signal is classified as ‘not applicable’ - if that makes sense. Generally, for this method, similar to the isovist , need to find out what the signal falloff point is. This would be translated in distance to which I’m trying to avoid for now as distance does not equal accuracy. Thoughts?