There are two main complexities to this build.

1. We require python-2.7s libraries
2. We need to build an old version of AI-toolbox because the current version requires C++ 17. We also require three C++ libraries to use AI-toolbox
3. We need to copy two shared libraries from AI-toolbox to PCog

This guide requires an ubuntu 16.04 box and python-2.7 installed. It also assumes that you have installed the QCog javacode base and the unity3d test bed.

This command should install all packages required to build and run PCog.

sudo apt-get update
sudo apt-get install libeigen3-dev liblpsolve55-dev python-tk graphviz python-pygraphviz libboost-all-dev build-essential

In this section we build AI-toolbox so that we can use its shared libraries to run PCog.

First clone AI-toolbox

git clone https://github.com/Ivan1931/AI-Toolbox.git

Now build the libraries:

cd AI-toolbox
git checkout origin/c++14-build
mkdir build
cd build
cmake ..
make

Now you should have the shared libraries located in build directory.

First clone PCog

git clone https://github.com/Ivan1931/pcog.git

Copy the build libraries from AI-toolbox to the deps directory in PCog.

cd pcog
cp /path/to/ai-toolbox/build/*.so pcog/deps

If all steps have been setup you can run PCog.

Install the necessary python libraries:

pip install -r .

Finally, run PCog

python -m pcog.__main__

To run the agent you first have to start the test bed, then start PCog, then start the java agent with the following command:

gradle run -Ppcog