Also known as the new version of the "RobotBrunch" codebase from HEATlab.

This code is closely tied to the paper:

Abrahams et al. 2019. DREAM: An Algorithm for Mitigating the Overhead of Robust Rescheduling. In Proc. of the 29th International Conference on Automated Planning and Scheduling (ICAPS-2019). A pre-print is available here.

DREAM now requires the COIN-OR CBC program, located at /usr/bin/cbc. It can be installed on Debian systems via

sudo apt install coinor-cbc

Requirements (python packages) to run are listed in requirements.txt.

These can be installed via your standard python3 pip method of pip install -r requirements.txt.

Newer versions may be used, but these versions have been selected to work on HMC's shared server, Knuth.

Usage

usage: run_simulator.py [-h] [-v] [-t THREADS] [-s SAMPLES] [-e EXECUTION]
                        [-o OUTPUT] [--ar-threshold AR_THRESHOLD]
                        [--si-threshold SI_THRESHOLD] [--mit-parse]
                        [--seed SEED] [--ordering-pairs ORDERING_PAIRS]
                        [--start-point START_POINT] [--stop-point STOP_POINT]
                        [--no-live]
                        stns [stns ...]

You may always use the --help option to get a full print out of every option.

How to use DREAM on an PSTN:

$ python3 run_simulator.py -e arsi -s 100 test_data/two_agent_sync.json

The -e option sets the execution strategy, and the -s sets the number of samples to simulate.

To generate Sphinx autodoc documentation:

1. Go to docs
2. Run make html.
3. Open the index now generated at docs/build/html/index.html using your favourite web browser.

Docstrings follow Google Docstring Style, and we use the Sphinx Napolean add-on to convert between Google style to RST for Sphinx.

All problem instances used to evaluate DREAM are available in the problem_instances directory. They are parameterized into subdirectories by problem features, as described in Lund et al 2017

The ROVERS benchmark is available at http://groups.csail.mit.edu/mers/datasets/scheduling/. rovers_coordination.json contains the instances used to evaluate DREAM in the 2019 ICAPS paper.

Check out the IDEAS.md file to see what we (as a lab) have tried.

This project supports the unit test module. To run the tests in this project, use the following shell command.

$ cd <project_dir>
$ python3 -m unittest  # Auto searches for unit tests in <project_dir>/tests

PuLP may give a warning. You can implement Jordan's fix here in PuLP itself to remove it.

This code is copyright of Harvey Mudd College's HEATlab.

Rewrite Started by: Jordan R. Abrahams (HMC '19)

Original Written by:

• Jeb Brooks
• Emilia Reed
• Alexander Gruver
• Sam Dietrich
• Kyle Lund
• Scott Chow
• Jordan R. Abrahams
• Brenner Ryan
• David A. Chu
• Grace Diehl
• Marina Knittel
• Judy Lin
• Liam Lloyd
• James C. Boerkoel Jr.

Please contact Jordan R. Abrahams if you have any questions about the internals of the code, documentation, or any of the results gathered.