强化学习案例复现（1）--- MountainCar基于Q-learning

1 搭建环境

1.1 gym自带

import gym

# Create environment
env = gym.make("MountainCar-v0")

eposides = 10
for eq in range(eposides):
    obs = env.reset()
    done = False
    rewards = 0
    while not done:
        action = env.action_space.sample()
        obs, reward, done, action, info = env.step(action)
        env.render()
        rewards += reward
    print(rewards)

1.2 自行搭建（建议用该方法）

按照下文搭建MountainCar环境

往期文章：强化学习实践（三）基于gym搭建自己的环境（在gym0.26.2可运行）-CSDN博客

 

2.基于Q-learning的模型训练

import gym
import numpy as np

env = gym.make("GridWorld-v0")

# Q-Learning settings
LEARNING_RATE = 0.1 #学习率
DISCOUNT = 0.95  #奖励折扣系数
EPISODES = 100  #迭代次数

SHOW_EVERY = 1000

# Exploration settings
epsilon = 1  # not a constant, qoing to be decayed
START_EPSILON_DECAYING = 1
END_EPSILON_DECAYING = EPISODES//2
epsilon_decay_value = epsilon/(END_EPSILON_DECAYING - START_EPSILON_DECAYING)

DISCRETE_OS_SIZE = [20, 20]
discrete_os_win_size = (env.observation_space.high - env.observation_space.low) / DISCRETE_OS_SIZE

print(discrete_os_win_size)


def get_discrete_state(state):

    discrete_state = (state - env.observation_space.low)/discrete_os_win_size

   # discrete_state = np.array(state - env.observation_space.low, dtype=float) / discrete_os_win_size

    return tuple(discrete_state.astype(np.int64))  # we use this tuple to look up the 3 Q values for the available actions in the q-


q_table = np.random.uniform(low=-2, high=0, size=(DISCRETE_OS_SIZE + [env.action_space.n]))


for episode in range(EPISODES):
    state = env.reset()
    discrete_state = get_discrete_state(state)

    if episode % SHOW_EVERY == 0:
        render = True
        print(episode)
    else:
        render = False

    done = False
    while not done:
        if np.random.random() > epsilon:
            # Get action from Q table
            action = np.argmax(q_table[discrete_state])
        else:
            # Get random action
            action = np.random.randint(0, env.action_space.n)

        new_state, reward, done, _, c = env.step(action)
        new_discrete_state = get_discrete_state(new_state)

        # If simulation did not end yet after last step - update Q table
        if not done:
            # Maximum possible Q value in next step (for new state)
            max_future_q = np.max(q_table[new_discrete_state])
            # Current Q value (for current state and performed action)
            current_q = q_table[discrete_state + (action,)]
            # And here's our equation for a new Q value for current state and action
            new_q = (1 - LEARNING_RATE) * current_q + LEARNING_RATE * (reward + DISCOUNT * max_future_q)
            # Update Q table with new Q value
            q_table[discrete_state + (action,)] = new_q
            # Simulation ended (for any reson) - if goal position is achived - update Q value with reward directly

        elif new_state[0] >= env.goal_position:
            # q_table[discrete_state + (action,)] = reward
            q_table[discrete_state + (action,)] = 0
            print("we made it on episode {}".format(episode))

        discrete_state = new_discrete_state

        if render:
            env.render()

    # Decaying is being done every episode if episode number is within decaying range
    if END_EPSILON_DECAYING >= episode >= START_EPSILON_DECAYING:
        epsilon -= epsilon_decay_value

np.save("q_table.npy", arr=q_table)

env.close()

3.模型测试

import gym
import numpy as np


env = gym.make("GridWorld-v0")

# Q-Learning settings
LEARNING_RATE = 0.1
DISCOUNT = 0.95
EPISODES = 10

DISCRETE_OS_SIZE = [20, 20]
discrete_os_win_size = (env.observation_space.high - env.observation_space.low) / DISCRETE_OS_SIZE

def get_discrete_state(state):
    discrete_state = (state - env.observation_space.low)/discrete_os_win_size
    return tuple(discrete_state.astype(np.int64))  # we use this tuple to look up the 3 Q values for the available actions in the q-

q_table = np.load(file="q_table.npy")

for episode in range(EPISODES):
    state = env.reset()
    discrete_state = get_discrete_state(state)

    rewards = 0
    done = False
    while not done:
        # Get action from Q table
        action = np.argmax(q_table[discrete_state])
        new_state, reward, done, _, c = env.step(action)
        new_discrete_state = get_discrete_state(new_state)

        rewards += reward

        # If simulation did not end yet after last step - update Q table
        if done and new_state[0] >= env.goal_position:
            print("we made it on episode {}, rewards {}".format(episode, rewards))

        discrete_state = new_discrete_state
        env.render()

env.close()