SurakartaAI

This Repository build a Surakarta Game framework and implement MCTS (Monte Carlo Tree Search) technique to train AI. In the future, we will combine MCTS with Deep Q-Learning or other appropriate RL, ML, DRL technique.

How to Run?

Important: You should install C++ libtorch libaray first

1. git clone this Repository
2. cd SurakartaAI
3. mkdir build
4. cd build
5. cmake -DCMAKE_PREFIX_PATH=/path/to/yourlibtorch ..
6. cmake --build .
7. ./surakarta (--total=1000 --block=100 (default argument))

How to Modify Strategy?

• By deault, The two agent player and envir move according to two different strategy, MCTS and Eat First Greedy.
• Modify agent.h and choose the strategy you want in policy.h