simopt-competition-26

Run the experiment and plot the results

The commands below are for macOS or Linux. Start with a local checkout of this repository and Git installed; the project fetches SimOpt from GitHub during setup.

1. Install uv:

   curl -LsSf https://astral.sh/uv/install.sh | sh

   Restart your terminal so it picks up uv, then check the installation:

   uv --version

2. Enter the repository directory, replacing the path below with your checkout location. Install Python 3.13, the version in .python-version, and sync the project dependencies:

   cd /path/to/simopt-competition-26
   uv python install 3.13
   uv sync

   uv creates a .venv for the project. You do not need to activate it before using uv run. See the uv project guide for details.

3. Run the experiment from the repository directory:

   uv run simopt-competition-26

   This runs RandomSearch, DemoSolver from demo.py, and DASTRODF from dastrodf.py on the tele-operator scheduling problem, performs the post-replications, and plots all three mean, unnormalized progress curves together in plot.png in the current directory at 300 dpi. DemoSolver searches feasible neighbors by adding or removing one driver at one shift start, using 10 simulation replications per candidate and accepting the first improvement in randomized order. It stops when the budget runs out or a full neighborhood gives no improvement. All three solvers use the same experiment settings and replication budget. Plotting happens automatically at the end of the run. Running the command again replaces plot.png.

   The experiment settings are in the entry point: 10 macroreplications, 20 post-replications, and 20 post-replications for the initial and reference solutions. The default optimization budget is 1000 simulation replications, each covering nine simulated days, so the full experiment can take time. A plot file has been saved at plot.png.