AI for the game 2048.

Developed by: Shiwani Bisht (sb854) Ajay Gandhi (aag255) David Vakili (dv227)

See it in action here.

We developed this game based on the source code for 2048 and adapted existing AI framework to develop our own artificial intelligence agents that could interact with the game. These agents use a board evaluation function that we developed to evaluate the desirability of a certain board configuration.

You can try to run two random agents, an agent that uses DFS, Minimax (with alpha-beta pruning), or Expectiminimax.

We found that Expectimax works the best!

Note: there's an animation delay of 100ms to show the tiles moving. This causes a bit of lag when running the search algorithms. DFS and expectimax take a while to run due to the large branching factor.