# Last edited on 2023-02-21 23:18:23 by stolfi

Data from/for the loctel projet of J. Stolfi and Cid Souza, ~2001.

See {stmap.h} in the library  {${STOLFIHOME}/programs/c/JSLIBS/libstreetmap}
and  ${STOLFIHOME}/programs/c/MISC/loctel for software
that generates, uses, and visualizes this data.

There are there datasets corresponding to three urban areas:

  "bahia"    some small town in Bahia.
  "campinas" a section of Campinas, SP.
  "itajobi"  a section of Itajobi, SP.

In each directory there are the following files:

  "map.dat"  The street grid map provided by CPqD,
             in their SAGRE format.

  "map.rnt"  The street map converted to the format accepted
             by the {stmap.h} functions.  Derived from the
             "map.dat" file by the script {convert-sagre-dat.gawk}
             
  "pop.pgm"  A PGM image that gives the population density in the
             neighborhood of each point, estimated from the 
             street density.

  "usr.vdt"  A table giving the estimated demand of telephone use 
             at each vertex of the map.
               
Among other things, the script {convert-sagre-dat.gawk} shifts the
absolute SAGE coordinates to coordinates relative to some local origin.
It also breaks map edges into segments with a specified maximum length.
See {convert_all.sh} for the local origin and edge split limit of each
map.

The "campinas/map-n2.rnt" and "itajobi/map-n2.rnt" files for seem to be
older and smaller versions of the corresponding "map.rnt" files, with
split limit 20 metres instead of 15 metres.

The "pop.pgm" images and "usr.vdt" files were created from the "map.rnt"
data with the program 
{${STOLFIHOME}/programs/c/MISC/loctel/demography.c}

The program first computes a /vertex street measure/ {vms[vi]} for each map
vertex {vi}, defined as the total length of the half-edges of the map
that are incident to the vertex.

Then it estimates a /population density/ {pop(x,y)} (in people per
square meter) at each point {x,y} of the map, proportional to the
distance-weighted average street density (in meters per square meter)
around the point, where the weights are a Gaussian distribution with a
specified radius. This data, properly scaled, is written out to the
"pop.pgm" images.

The "usr.vdt" files give the estimated /demand/ at each vertex of the
map, computed as the vertex's street measure times the estimated population
density at that point.