It's been a while since I looked into this myself, but I used to do a lot of GIS work and have a lot of experience with the tools for doing this kind of analysi

It's been a while since I looked into this myself, but I used to do a lot of GIS work and have a lot of experience with the tools for doing this kind of analysis. Doing a quick search around for a general purpose algorithm for the "traveling salesman problem" - I came across this: http://www-sre.wu-wien.ac.at/ersa/ersaconfs/ersa03/cdrom/papers/433.pdf http://www-sre.wu-wien.ac.at/ersa/ersaconfs/ersa03/cdrom/pap... In my experience, the most important part of doing network analysis is to have a "topology model" for your road network data. It contains the information necessary to model turns, drive-time, traffic signals, etc. Anyway, I used to know the 1000 ft view of the math behind this stuff but not anymore. If, however, you're trying to build it into an application - there are solutions like pgRouting https://github.com/pgRouting/pgrouting https://github.com/pgRouting/pgrouting which utilizes PostGIS (PostgreSQL), ESRI's Network Analyst ($$), and Google Maps Directions API. These all leverage some underlying topology data like Open Street Map to do routing and figure out the subsequent times/schedule to get from point to point to point... Hope this helps!