while True:
# fresh env
env.render()
# RL choose action based on observation
action = RL.choose_action(str(observation))
# RL take action and get next observation and reward
observation_, reward, done = env.step(action)
# RL learn from this transition
RL.learn(str(observation), action, reward, str(observation_))
# swap observation
observation = observation_
# 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 = QLearningTable(actions=list(range(env.n_actions)))
env.after(100, update)
env.mainloop()
# RL choose action based on observation
action = RL.choose_action(str(observation))
# RL take action and get next observation and reward
observation_, reward, done = env.step(action)
# RL learn from this transition
RL.learn(str(observation), action, reward, str(observation_))
# swap observation
observation = observation_
# break while loop when end of this episode
if done:
break
# end of game
print('game over')
env.destroy()
if __name__ == "__main__":
env = Maze()
# import numpy as np
# print(np.arange(4))
# RL = QLearningTable(actions=list(np.arange(4)))
RL = QLearningTable(actions=list(range(env.n_actions)))
env.after(100, update)
env.mainloop()
experiments = []
# # alg0 (Aynsc)
# env0 = Maze(agentXY,goalXY,wall_shape, pits)
# RL0 = rlalg0(actions=list(range(env0.n_actions)))
# data0={}
# env0.after(10, update(env0, RL0, data0, episodes))
# env0.mainloop()
# experiments = [(env0,RL0, data0)]
# alg2 (SARSA)
env2 = Maze(agentXY, goalXY, wall_shape, pits)
RL2 = rlalg2(actions=list(range(env2.n_actions)))
data2 = {}
env2.after(10, update(env2, RL2, data2, episodes))
env2.mainloop()
experiments.append((env2, RL2, data2))
# alg1 (Q-Learning)
env1 = Maze(agentXY, goalXY, wall_shape, pits)
RL1 = rlalg1(actions=list(range(env1.n_actions)))
data1 = {}
env1.after(10, update(env1, RL1, data1, episodes))
env1.mainloop()
experiments.append((env1, RL1, data1))
# alg4 (Expected Sarsa)
env4 = Maze(agentXY, goalXY, wall_shape, pits)
RL4 = rlalg4(actions=list(range(env4.n_actions)))
data4 = {}
env4.after(10, update(env4, RL4, data4, episodes))
# break while loop when end of this episode
if done:
break
step += 1
# end of game
print('game over')
env.destroy()
if __name__ == "__main__":
# maze game
env = Maze()
# RL方法选择DQN
RL = DeepQNetwork(env.n_actions, env.n_features,
learning_rate=0.01, # 学习效率设为0.01（）
reward_decay=0.9, # 预计回报衰减
e_greedy=0.9, # 选择最大Q值对应的动作的概率
replace_target_iter=200, # 每隔200步替换一次target_net的参数
memory_size=2000, # 记忆上限
output_graph=True, # 输出神经网络训练模型
restore_network=False,
save_network=False
)
RL.restore_net()
env.after(100, run_maze) # after语句可以实现定时器循环
env.mainloop() # mainloop就进入到事件（消息）循环
save_path = RL.save_net()
RL.plot_cost() # 观看神经网络的误差曲线
# reinforcement learning
if (step > 5) and (step % 5 == 0):
RL.learn()
# take action
observation = observation_
if done: break
step += 1
print('game over')
env.destroy()
if __name__ == "__main__":
# maze game
env = Maze()
RL = DeepQNetwork(env.n_actions,
env.n_features,
learning_rate=0.01,
reward_decay=0.9,
e_greedy=0.9,
replace_target_iter=200,
memory_size=2000)
# output_graph=True)
env.after(100, run_maze) # wait for initialize
env.mainloop() # start environment
RL.plot_cost()
[3, 6], [4, 6], [5, 6]])
pits = []
if Task == "T3":
#Task 3
wall_shape = np.array([[7, 4], [7, 3], [6, 3], [6, 2], [5, 2], [4, 2],
[3, 2], [3, 3], [3, 4], [3, 5], [3, 6], [4, 6],
[5, 6]])
pits = np.array([[1, 3], [0, 5], [7, 7]])
# sarsa
env1 = Maze(agentXY, goalXY, wall_shape, pits)
RL1 = rlalg1(actions=list(range(env1.n_actions)))
data1 = {}
env1.after(10, update(env1, RL1, data1, episodes))
env1.mainloop()
experiments = [(env1, RL1, data1)]
# Q-learning
env2 = Maze(agentXY, goalXY, wall_shape, pits)
RL2 = rlalg2(actions=list(range(env2.n_actions)))
data2 = {}
env2.after(10, update(env2, RL2, data2, episodes))
env2.mainloop()
experiments.append((env2, RL2, data2))
# Expected Sarsa
env3 = Maze(agentXY, goalXY, wall_shape, pits)
RL3 = rlalg3(actions=list(range(env3.n_actions)))
data3 = {}
env3.after(10, update(env3, RL3, data3, episodes))