My implementation of online 2D irregular BPP algorithm with state-of-the-art picking policy minimal surrounding waste.
Licence: The software as a whole is licenced under GPL v3 because I'm using GPL v3-licenced library libnfporb, but for the everything except repository source/nfp_interface I'm granting MIT licence.
+I would be happy if you wrote to me when you use my project because I'm curious if this will be useful for someone :)
Note: If you want to replace libnfporb library with your own implementation of No-fit polygon and Inner-fit polygon calculation, then you need to modify only the geometry_tools.py file (call your own functions, maybe modify error handling, ...) to make it work.
Build: see Requirements.md for instructions.
Usage: run run_tests.py for a sample computation. If you choose best_run, then it runs on dataset generators from SFG_competition using specified parameters. euro_datasets loads and runs specified datasets from .csv files.
Versions:
Current version uses last version of libnfporb (the library was archived on 8th March 2023) with my modifications and requires boost version 1.76.0. It's more robust than older version => less errors in NPF computations => very slightly different results and significantly faster times on some of the EURO datasets (those with a lot of NFP computation errors).
For latest version that uses original libnfporb library with boost version 1.65.0, see tag last-boost-1.65.0. Results included in results folder should be reproducible with this version, at least on Windows.
Performance note: it seems that when building the libnfprob with cmake (at least on Linux), there is a big (around 10x) performance penalty compared to building with Visual Studio (with average time to place 1 shape (using tests in run_test.py) around +- 1 second). If you figure out how to compile it more effectively, please let me know.
References: My Bachelor thesis (written in Czech).