TBG stands for Topology Based Geolocation.
Findings:
Fida generated datasets through TBG code for various landmarks. Each landmark has hundreds of estimates for its possible geolocation. All of these estimates (for each landmark) are put into a file (estimates.out). TBG by default takes the first entry from each file and calculates the error against its (known) original position. Fida took other estimates for landmarks and found that other values can potentially reduce the error. So the problem statement was to find which value can yield minimum possible error and maximum possible accuracy without knowing the original position of the landmark.
Assumption:
We do not consider or use the actual location in any of the calculations, rather we try to find which estimate yields maximum number of other estimates in its vicinity.
Method:
To find this we are using the following technique:
- Calculate distance from each estimate to all other estimates of a target. This is repeated for every single estimate of a target.
- Filter estimates which are less than 250km away from other estimates for each target.
- Filter those lists which contain maximum number of estimates of each target.
- Filter those results which contain the best estimate. The best estimate is the estimate found by Fida to reveal the least possible error for location of a target (among all estimates for a target).
- Calculate arithmetic mean or median and see if it's anywhere near the best estimate.
Purpose:
The aim to do all this is to find whether a cluster forms around the best estimate for geolocation of a landmark. This may potentially enable us to find a landmark with certain amount of accuracy without knowing original position or best estimate.
Results:
This technique has failed. Spreadsheet here <https://confluence.slac.stanford.edu/download/attachments/88310312/TBG_improved_results_v1.xlsx>.