Multi agent simulator where multiple agents learn the behvior of an unknown adversary using language theory constructs (strictly-2-piecewise assumption). The software is mostly built using python. Matlab's python engine was intially used for supporting some opertions. Later this was fully replaced by python to offer better portability.

This work was supported by NSF CNS under awards #103588 and #103577. This project was a collaboration between University of Delaware and Boston university.

Sections of the code related to policy synthesis was written by Kevin Leahy of Boston University.

The FSA package was used for visualization purposes. This was later superseded by a custom build of FSA functions that were developed expressly for this application.

For more information about the FSA package see https://pypi.python.org/pypi/FSA

If you use as part of this package, please cite Kevin Leahy, Prasanna Kannappan, Adam Jardine, Jeffrey Heinz, Bert Tanner, and Calin Belta, "Integration of Deterministic Inference with Formal Synthesis for Control under Uncertainty.", submitted to American Control Conference (ACC) 2015.

This package was developed using python 2.7 64-bit distribution. The testing was done in ubuntu 12.04. The software should be portable to other operating systems (e.g. mac, windows) if the appropriate python distribution is available. However, this has not been verified. Let me know if you have encounter any problems.

There are two executable files in the gui folder. To run any executable file in linux, open a terminal window and point to the gui folder. You can use the the cd command to change directory.

e.g. /home/cps_multi_agent/gui

Once you have changed the directory to the gui directory, then you can run the executables. For example, o run the cpsGameViewer.py executable, use the command below.

./cpsGameViewer.py

If the file does not have executable permissions, then use

python cpsGameViewer.py

If your default version of python is not 2.7 then use

python2.7 cpsGameViewer.py

1. cpsGameGTS.py

This executable can be used to start a new multi-agent game. There are several gui controls that can be used to set the parameters of the game. This simulation can involve an execution time of a few minutes to few hours based on the parameters of the game. Most of the gui controls are meant to be straight-forward and easy to use.

2. cpsGameViewer.py

This executable lets you run a pre-recorded game (recording done using cpsGameGTS.py). It lets you choose a game log file using gui controls and allows walking through the game moves sequentially. Some recorded games can be found in the game_runs folder. There is also a video game_recording.webm showing the recorded game as displayed by cpsGameViewer. The gui controls are self explanatory.

1. Missing Dependencies

All the libraries used here can be easily downloaded from python repositories. In case you have any missing imports. Try installing them using pip.

e.g. sudo pip install missing_dependence

2. Changing permissions of a file to executable

Use command:

sudo chmod 755 file_name