travisddavies / pacman_mdp_and_reinforcement_learning_algorithms

Various algorithms for Markov's Decision Process and Reinforcement Learning in a grid world and in Pacman, including Value Iteration, Q Learning and Approximate Q-Learning

pacman_mdp_and_reinforcement_learning_algorithms

Various algorithms for Markov's Decision Process and Reinforcement Learning in a grid world and in Pacman, including Value Iteration, Q Learning and Approximate Q-Learning.All code is provided by UC Berkeley, apart from the algorithms in question.

Value Iteration

To play around with value iteration for Markov's Decision Process, you can use the following commands.

python gridworld.py -a value -i 100 -k 10
python gridworld.py -a value -i 5

Q-Learning

To play around with Q-learning for Markov's Decision Process, you can use the following commands.

python gridworld.py -a q -k 5 -m

Epsilon Greedy

To play around with Epsilon Greedy for Markov's Decision Process, you can use the following commands.

python gridworld.py -a q -k 100
python gridworld.py -a q -k 100 --noise 0.0 -e 0.1
python gridworld.py -a q -k 100 --noise 0.0 -e 0.9

To play around with a crawler learning to walk, you can use the following commands.

python crawler.py

Q-Learning and Pacman

To play around with PacMan using Q-Learning, you can use the following commands.

python pacman.py -p PacmanQAgent -x 2000 -n 2010 -l smallGrid
python pacman.py -p PacmanQAgent -n 10 -l smallGrid -a numTraining=10

Approximate Q-Learning

To play around with training pacman with Approximate Q-Learning, you can use the following commands.

python pacman.py -p ApproximateQAgent -a extractor=SimpleExtractor -x 50 -n 60 -l mediumGrid
python pacman.py -p ApproximateQAgent -a extractor=SimpleExtractor -x 50 -n 60 -l mediumClassic