Notes

#simple q learning env using openai gym(Frozenlake)

import gym
import numpy as np
def update_Q(Q, state, action, reward, next_state, alpha, gamma):
current_Q = Q[state, action]
max_next_Q = np.max(Q[next_state])
new_Q = current_Q + alpha * (reward + gamma * max_next_Q - current_Q)
Q[state, action] = new_Q
return Q

env = gym.make('FrozenLake-v1')
n_episodes = 10000
max_steps_per_episode = 100

alpha = 0.8
gamma = 0.95
epsilon = 0.2

Q = np.zeros((env.observation_space.n, env.action_space.n))

for episode in range(n_episodes):
state = env.reset()
done = False
step = 0

while not done and step < max_steps_per_episode:
# choose action using epsilon-greedy policy
if np.random.uniform() < epsilon:
action = env.action_space.sample()
else:
action = np.argmax(Q[state])

# take action and observe the resulting state and reward
next_state, reward, done, info = env.step(action)

# update Q-values
Q = update_Q(Q, state, action, reward, next_state, alpha, gamma)

# transition to the next state
state = next_state
step += 1

env.close()