A minimum working example for using the SOAR cognitive architecture with python for creating an agent that interacts with the environment using SOAR's input-output links.
On a linux machine:
-
Install java OpenJDK v8
-
Install python 2 and python 3 2.1) Make python 3 your default option by adding this to .bashrc:
alias python=python3
-
Install swig
sudo apt install swig
-
Clone Soar from github:
git clone https://github.com/SoarGroup/Soar.git
-
At Soar's root folder, run python build script:
If you want to use python 2, run this:
python scons/scons.py sml_python
However, if you want to use python 3, run the following:
python2 scons/scons.py sml_python --python=/usr/bin/python3
Build results should be in folder /out
At a terminal, export LD_LIBRARY_PATH and PYTHONPATH to Soar's code current location, for instance:
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:~/Desktop/Soar/out
export PYTHONPATH=$PYTHONPATH:~/Desktop/Soar/out
An alternative is to copy those lines to the end of .bashrc file, so you don't have to do it every time you open a new terminal.
Run hello world example:
python hello-world.py
You should see the following:
Hello world
1: ==>S: S2 (state no-change)
Run the toy environmnet example:
python toy-env.py
You should see something like this:
(' ------------- Soar cycle: ', 0, ' ------------- ')
('---> Environment sensed: ', [7, 9])
===> Soar output: 16.000000
===> Soar input: 7.000000 9.000000
1: O: O2 (initialize-toy-env)
('---> Environment acted:', 0)
(' ------------- Soar cycle: ', 1, ' ------------- ')
('---> Environment sensed: ', [7, 3])
===> Soar output: 10.000000
===> Soar input: 7.000000 3.000000
2: ==>S: S2 (state no-change)
('---> Environment acted:', 16.0)
(' ------------- Soar cycle: ', 2, ' ------------- ')
('---> Environment sensed: ', [3, 4])
===> Soar output: 7.000000
===> Soar input: 3.000000 4.000000
3: ==>S: S3 (state no-change)
('---> Environment acted:', 10.0)
soar>
The goal of this exercise is to show how to use the IO interface and perform a simple cognitive cycle inside a Soar agent. The "toy environment" is just a function that produces two random numbers and receives a single result. This agent receives these numbers through "sensors", add them up, and return them through actuators.
-
Soar Manual http://soar.eecs.umich.edu/downloads/Documentation/SoarManual.pdf
-
Soar Tutorials https://soar.eecs.umich.edu/articles/downloads/soar-suite/228-soar-tutorial-9-6-0
-
Python Interface Example https://soar.eecs.umich.edu/articles/downloads/examples-and-unsupported/183-python-interface-example
-
Python SML Interface file /Soar/build/Core/ClientSMLSWIG/Python/Python_sml_ClientInterface.py
-
pysoarlib https://github.com/amininger/pysoarlib