莫烦python强化学习系列－－Sarsa-lambda学习

        Sarsa 是一种单步更新法, 在环境中每走一步, 更新一次自己的行为准则, 我们可以在这样的 Sarsa 后面打一个括号, 说他是 Sarsa(0), 因为他等走完这一步以后直接更新行为准则. 如果延续这种想法, 走完这步, 再走一步, 然后再更新, 我们可以叫他 Sarsa(1). 同理, 如果等待回合完毕我们一次性再更新呢, 比如这回合我们走了 n 步, 那我们就叫 Sarsa(n). 
        为了统一这样的流程, 我们就有了一个 lambda 值来代替我们想要选择的步数, 这也就是 Sarsa(lambda) 的由来.

        其实 lambda 就是一个衰变值, 他可以让你知道离奖励越远的步可能并不是让你最快拿到奖励的步, 所以我们想象我们站在宝藏的位置, 回头看看我们走过的寻宝之路, 离宝藏越近的脚印越看得清, 远处的脚印太渺小, 我们都很难看清, 那我们就索性记下离宝藏越近的脚印越重要, 越需要被好好的更新. 和之前我们提到过的奖励衰减值 gamma 一样, lambda 是脚步衰减值, 都是一个在 0 和 1 之间的数.

        当 lambda 取0, 就变成了 Sarsa 的单步更新, 
        当 lambda 取 1, 就变成了回合更新, 对所有步更新的力度都是一样. 
        当 lambda 在 0 和 1 之间, 取值越大, 离宝藏越近的步更新力度越大. 

        具体算法流程见下图：

         
具体代码

from maze_env import Maze
from RL_brain import SarsaLambdaTable


def update():
    for episode in range(100):
        # initial observation
        observation = env.reset()

        # RL choose action based on observation
        action = RL.choose_action(str(observation))

        # initial all zero eligibility trace
        # 新回合, 清零
        RL.eligibility_trace *= 0

        while True:
            # fresh env
            env.render()

            # RL take action and get next observation and reward
            observation_, reward, done = env.step(action)

            # RL choose action based on next observation
            action_ = RL.choose_action(str(observation_))

            # RL learn from this transition (s, a, r, s, a) ==> Sarsa
            RL.learn(str(observation), action, reward, str(observation_), action_)

            # swap observation and action
            observation = observation_
            action = action_

            # break while loop when end of this episode
            if done:
                break

    # end of game
    print('game over')
    env.destroy()

if __name__ == "__main__":
    env = Maze()
    RL = SarsaLambdaTable(actions=list(range(env.n_actions)))

    env.after(100, update)
    env.mainloop()

import numpy as np
import pandas as pd


class RL(object):
    def __init__(self, action_space, learning_rate=0.01, reward_decay=0.9, 
                 e_greedy=0.9):
        self.actions = action_space  # a list
        self.lr = learning_rate
        self.gamma = reward_decay
        self.epsilon = e_greedy

        self.q_table = pd.DataFrame(columns=self.actions, dtype=np.float64)

    def check_state_exist(self, state):
        if state not in self.q_table.index:
            # append new state to q table
            self.q_table = self.q_table.append(
                pd.Series(
                    [0]*len(self.actions),
                    index=self.q_table.columns,
                    name=state,
                )
            )

    def choose_action(self, observation):
        self.check_state_exist(observation)
        # action selection
        if np.random.rand() < self.epsilon:
            # choose best action
            state_action = self.q_table.loc[observation, :]
            # some actions may have the same value, 
            # randomly choose on in these actions
            action = np.random.choice(state_action[state_action == 
                                      np.max(state_action)].index)
        else:
            # choose random action
            action = np.random.choice(self.actions)
        return action

    def learn(self, *args):
        pass


# backward eligibility traces
class SarsaLambdaTable(RL):
    def __init__(self, actions, learning_rate=0.01, reward_decay=0.9, e_greedy=0.9, 
                 trace_decay=0.9):
        super(SarsaLambdaTable, self).__init__(actions, learning_rate, reward_decay, 
                                               e_greedy)

        # backward view, eligibility trace.
        self.lambda_ = trace_decay
        self.eligibility_trace = self.q_table.copy()

    def check_state_exist(self, state):
        if state not in self.q_table.index:
            # append new state to q table
            to_be_append = pd.Series(
                    [0] * len(self.actions),
                    index=self.q_table.columns,
                    name=state,
                )
            self.q_table = self.q_table.append(to_be_append)

            # also update eligibility trace
            self.eligibility_trace = self.eligibility_trace.append(to_be_append)

    def learn(self, s, a, r, s_, a_):
        self.check_state_exist(s_)
        q_predict = self.q_table.loc[s, a]
        if s_ != 'terminal':
            q_target = r + self.gamma * self.q_table.loc[s_, a_]  
            # next state is not terminal
        else:
            q_target = r  # next state is terminal
        error = q_target - q_predict

        # increase trace amount for visited state-action pair

        # Method 1:
        # 对于经历过的 state-action, 我们让他+1, 证明他是得到 reward 路途中不可或缺的一环
        # self.eligibility_trace.loc[s, a] += 1

        # Method 2:
        self.eligibility_trace.loc[s, :] *= 0
        self.eligibility_trace.loc[s, a] = 1

        # Q update
        self.q_table += self.lr * error * self.eligibility_trace

        # decay eligibility trace after update
        self.eligibility_trace *= self.gamma*self.lambda_