Fix: replaced borked location prediction
The location prediction fails bigtime in previous versions,
because of wrong geometrical assumptions. Tried to correct
this in this commit. Prediction now depends on the number of
towers that are known at the given moment:
0: no prediction
1: location is that of the known tower
2: between the two known towers, closer to the higher signal
3: triangulate for three towers
>3: triangulate for all possible triads, with the restriction
that the connected tower has to be among the three (the
other two changes). Check distance of predictions from
connected tower, throw away the far half, average the
positions of the rest.
if the connected tower is not among the known towers,
find the one with the highest signal and pretend that
is the connected tower. Do triangulations as otherwise.
Tested at a stationary position position (at my desk;) and
seemed to work but going to try in motion. This algorithm
needs to be tested a lot....