Danaze

import numpy as np
import gym
env = gym.make("CartPole-v1")
# there are two actions 
print("number of actions: " + np.str(env.action_space.n))
# check the observations bounds for each (position, velocity, angle, angular velocity)
for _ in range(4):
    print (env.observation_space.high[_])
    print (env.observation_space.low[_])

Danaze

    def discretize_state(self, obs):
        discretized = list()
        for i in range(len(obs)):
            scaling = (obs[i] + abs(self.lower_bounds[i])) / (self.upper_bounds[i] - self.lower_bounds[i])
            new_obs = int(round((self.buckets[i] - 1) * scaling))
            new_obs = min(self.buckets[i] - 1, max(0, new_obs))
            discretized.append(new_obs)
        return tuple(discretized)

Danaze

    def QLupdate(self, state, action, reward, new_state):
        # updating the Q-value of the visited state-action pair
        self.Q_table[state][action] += self.learning_rate * (reward + self.discount * np.max(self.Q_table[new_state]) - self.Q_table[state][action])
    def SARSAupdate(self, state, action, reward, new_state, next_action):
        # updating the Q-value of the visited state-action pair
        self.Q_table[state][action] += self.learning_rate * (reward + self.discount * self.Q_table[new_state][next_action] - self.Q_table[state][action])

Danaze

    def QLtrain(self):
        cum_reward = np.zeros((self.num_episodes))
        for ep in range(self.num_episodes):
             current_state = self.discretize_state(self.env.reset())
            done = False
            while not done:
                #choosing action according to our exploration-exploitation policy
                action = self.choose_action(current_state)
                obs, reward, done, _ = self.env.step(action)
                cum_reward[ep]+=reward