Pacman RL agent

Code files edited: ./search/search.py ./multiagent/multiAgents.py ./reinforcement/qlearningAgents.py

Search algorithms: [requires python2 environment to run following commands]

DFS

Tiny - python pacman.py -l tinyMaze -p SearchAgent Medium - python pacman.py -l mediumMaze -p SearchAgent --frameTime=0.01 Big - python pacman.py -l bigMaze -z .5 -p SearchAgent --frameTime=0.01

BFS

Tiny - python pacman.py -l tinyMaze -p SearchAgent -a fn=bfs Medium - python pacman.py -l mediumMaze -p SearchAgent -a fn=bfs --frameTime=0.01 Big - python pacman.py -l bigMaze -z .5 -p SearchAgent -a fn=bfs --frameTime=0.01

Uniform cost search

Tiny - python pacman.py -l tinyMaze -p SearchAgent -a fn=ucs Medium - python pacman.py -l mediumMaze -p SearchAgent -a fn=ucs --frameTime=0.01 Big - python pacman.py -l bigMaze -z .5 -p SearchAgent -a fn=ucs --frameTime=0.01 Medium Dotted - python pacman.py -l mediumDottedMaze -p StayEastSearchAgent Medium Scary - python pacman.py -l mediumScaryMaze -p StayWestSearchAgent

A-star - Heuristic

Tiny - python pacman.py -l tinyMaze -p SearchAgent -a fn=astar,heuristic=manhattanHeuristic Medium - python pacman.py -l mediumMaze -p SearchAgent -a fn=astar,heuristic=manhattanHeuristic --frameTime=0.01 Big - python pacman.py -l bigMaze -z .5 -p SearchAgent -a fn=astar,heuristic=manhattanHeuristic --frameTime=0.01

A-star - Without Heuristic

Tiny - python pacman.py -l tinyMaze -p SearchAgent -a fn=astar,heuristic=nullHeuristic Medium - python pacman.py -l mediumMaze -p SearchAgent -a fn=astar,heuristic=nullHeuristic --frameTime=0.01 Big - python pacman.py -l bigMaze -z .5 -p SearchAgent -a fn=astar,heuristic=nullHeuristic --frameTime=0.01

Multi-agent search algorithms: [Pacman will play looking ahead depth=3 steps.]

Minimax search

small - python pacman.py -p MinimaxAgent -a depth=4 -l smallClassic medium - python pacman.py -p MinimaxAgent -a depth=4 -l mediumClassic

Alpha-beta pruning

small - python pacman.py -p AlphaBetaAgent -a depth=3 -l smallClassic medium - python pacman.py -p AlphaBetaAgent -a depth=3 -l mediumClassic

Reinforcement learning:

[requires python3 environment to run following commands]

Q-learning

small - python pacman.py -p PacmanQAgent -x 1400 -n 1410 -l smallGrid --frameTime=0.1 -a epsilon=0.05 medium - python pacman.py -p PacmanQAgent -x 2000 -n 2010 -l mediumClassic --frameTime=0.1 -a epsilon=0.05

Approximate Q-learning

small - python pacman.py -p ApproximateQAgent -a extractor=SimpleExtractor -x 50 -n 60 -l smallGrid medium - python pacman.py -p ApproximateQAgent -a extractor=SimpleExtractor -x 50 -n 60 -l mediumGrid mediumClassic - python pacman.py -p ApproximateQAgent -a extractor=SimpleExtractor -x 0 -n 10 -l mediumClassic --frameTime=0.01