SoarGroup/Domains-InfiniteMario

This environment is derived from the Infinite Mario domain from RLCompetition2009 (based on RL-Glue) and uses SML to connect to Soar. Agents play a variant of Super Mario, a complete side-scrolling video game with destructible blocks, enemies, fireballs, coins, chasms, platforms, etc. The state space is complicated, but factored in an object-oriented way, which captures many aspects of the real world.

Running Instructions

1. Download and install Soar.
2. Configure environment variables for Soar ($SOAR_HOME)
3. Perform the install instructions found at RL-Competition 2009 software. The software is already in the download within the 15-rl-competition-2009 directory
4. Configure $COMP_HOME to point to the topmost directory of your local install of the competition software
5. Configure $AGENT_HOME to point to the topmost directory of your local install of MarioSoar
6. cd $AGENT_HOME
7. make clean;make
8. Run the agent ./run.bash config/combined.config
9. cd to trainer $COMP_HOME/trainers/guiTrainerJava/ for GUI trainer and ./run.bash or $COMP_HOME/trainers/consoleTrainerJava/ for headless trainer

Related Publications

Shiwali Mohan and John Laird. An Object-Oriented Approach to Reinforcement Learning in an Action Game. In Proceedings of 7th the Artificial Intelligence for Interactive Digital Entertainment Conference, AIIDE, 2011.

Shiwali Mohan and John Laird. Relational Reinforcement Learning in Infinite Mario. In Proceedings of the 24th AAAI Conference on Artificial Intelligence, AAAI, 2010. (Extended Abstract).

Shiwali Mohan and John E. Laird. Learning to Play Mario. Technical Report CCA-TR-2009-03, Center for Cognitive Architecture, University of Michigan, Ann Arbor, Michigan, 2009.