loggi-osrm research:

Inside data/

A sample from loggibud's data

Clustering

Since the data is too big to be handled in a single go, I decided to break the problem into smaller clusters that can be stored in memory, thus reducing the number of HTML requests made.

I'm no specialist in data mining, so I did some test runs in sklearn and ran the following clustering algorithms

DBScan is poorly optimized, but I tried plenty of parameters, and even the best one didn't manage to cluster the with minimum quality.

Inside src/

Still the first attempt. Did not work because the OSRM docker crashes after a fixed number of requests. Decided to fix it by clustering points into regions, but I should try to run the OSRM C API too, later.

cpp files:

• bad-opframe-main.cpp has all the necessary pieces from optframe, but the variables aren't properly isolated. Should be used as a template for further improvement.

• curlpp-example.cpp is reference I use to remember how to use curlpp and osrm http requests

• main-loggi.cpp was used to test the different algorithms

• optframe-main.cpp is used to call optframe and actually try to optimize the problem.

hpp files:

• distances-storage.hpp contains the storage of distances in a matrix, as to avoid http requests. I should include said function in context-loader.hpp, but it's hard to compile it, because it needs more structures that also need context-loader.hpp

• FCore_functions.hpp have the functions that will be used in OptFrame's FCore. Needs a refactor!

• loggi-context-loader.hpp has all the main class definitions regarding the problem itself.

• loggi-evaluator.hpp has functions that call OSRM to evalute distances

• loggi-include.hpp calls necessary functions and every other hpp in the correct order.

• loggi-solution-generators.hpp has the initial solutions. Currently a dumb one that generates on the giver order of the .json file and another that was based on the puca interpretation of the Clarke-Wright algorithm

• MoveData.hpp has the template that will be used to specify the Move's position

• NSSeq2Opt.hpp id the first Move implemented, currently being used to test if the integration is running smoothly.