代码实现Q-learning

深度学习Q—learning Q矩阵的更新基本公式如下：
Q_new（state，action）=(1-alpha)Q(state,action)+
alpha(R(state,action)+gamma*max*Q(stae_next,action_next))
以10X10矩阵为例代码实现为：

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
#reward = pd.read_excel("Qdata.xls")#如果以excel形式导入数据，可以使用这两个语句
#reward = np.array(reward)
#根据场景需求定义初始R矩阵
reward = np.array([[-1, 0, -1, -1, 0, -1, -1, -1, -1, -1],
                   [0, -1, -1, -1, -1, 0, -1, -1, -1, -1],
                   [-1, -1, -1, -1, -1, -1, 0, -1, -1, 100],
                   [-1, -1, -1, -1, -1, -1, 0, -1, -1, -1],
                   [0, -1, -1, -1, -1, 0, -1, -1, -1, -1],
                   [0, 0, -1, -1, 0, -1, 0, 0, -1, -1],
                   [-1, -1, 0, 0, -1, 0, -1, 0, -1, -1],
                   [-1, -1, -1, -1, -1, 0, 0, -1, 0, -1],
                   [-1, -1, -1, -1, -1, -1, -1, 0, -1, 100],
                   [-1, -1, 0, -1, -1, -1, -1, -1, 0, 100]])

Q_matrix = np.zeros((len(reward), len(reward)))#生成初始Q矩阵
rows, cols = reward.shape#生成state，action的行列个数
print(rows,cols)
steps = 0
gamma = 0.8
while steps < 500:
    steps += 1
    start_state = np.random.randint(0, rows)#随机选择一个state
    Rmax = max(reward[start_state])#得到最大的R矩阵
    for i in range(cols):
        if reward[start_state, i] != -1: #判断并不选择reward=-1的
            maxQ = max(Q_matrix[i])#选择下个action的最大reward
            #得到Q矩阵，并迭代循环
            Q_matrix[start_state, i] = reward[start_state, i] + gamma * maxQ 

print(np.round(Q_matrix/5))

原理参考网址：
https://blog.csdn.net/wei1033701020/article/details/54311113