Replanning in Advance for Instant Delay Recovery

This project has the following directories:

• generation: Python module to generate the @SIPP search graph
• search (atSIPP): C++ module to search for any-start-time plans in the @SIPP search graph
• data: two dutch shunting yard layouts: Enkhuizen and Heerlen. This also includes code to generate new scenarios and explanation of how the real-life scenario was created.
• experiments: the notebook contains all the code to run experiments for our paper

Dependencies (version tested):

• gcc (13.2.1)
• boost (1.83)
• zlib (1.3.1)
• meson (1.2.3)

Additionally, the Python generation module requires the numpy package to be installed, we tested using version 1.25.1.

Compiling:

    cd search
    meson setup --buildtype release  build
    meson compile -C build
    meson setup --buildtype debug build_debug
    meson compile -C build_debug

To run a specific scenario (in this case scenario data/enkhuizen/scenario_small_custom.json on location data/enkhuizen/location_enkhuizen.json for agent 1):

python3 generation/generate.py -s data/enkhuizen/scenario_small_custom.json -l data/enkhuizen/location_enkhuizen.json -o output
./search/build/atsipp --edgegraph output --start t-405B --goal t-401A

To cite, please use:

Issa Hanou, Devin W. Thomas, Wheeler Ruml, and Mathijs de Weerdt. Replanning in Advance for Instant Delay Recovery in Multi-Agent Applications: Rerouting Trains in a Railway Hub. (2024). In Proceedings: International Conference on Automated Planning and Scheduling.