def __init__(self, tau=True):
self.action_num = 4
self.shape = [7, 7]
self.time = 0
self.state = self.reset()
self.components = 7
_, state_num = np.where(self.state[:, 0:6] == 1)
self.state_num = state_num
if tau:
plot_state(self.state, X=7, Y=7)
def randomint(self, stau=False):
s = np.random.randint(0, 6, 7)
self.time = np.random.randint(0, 100, 1)[0]
raw_s = np.zeros([7, 7], dtype=np.int32)
raw_s[:, 0:6] = processsa(s, 6)
raw_s[:, 6] = time_encoder(self.time)
if stau == True:
plot_state(raw_s, 7, 7)
self.state = raw_s
self.state_num = s
return self.state, self.time
actions = np.zeros([num_episodes,100,7], dtype=np.int32)
states = np.zeros([num_episodes,101,7], dtype=np.int32)
path1 = simulation_path + '/training-step' + str(step)
if os.path.exists(path1)==False:
os.mkdir(path1)
for i_episode in range(num_episodes):
t=0
rAll = 0
done = False
current_state = env.reset()
a = []
states[i_episode,0] = env.state_num
while not done:
plt.figure(1)
plot_state(current_state, 7, 7)
plt.xlabel('states', **font)
plt.ylabel('components', **font)
state = np.reshape(current_state,[49])
action = sess.run(mainQN.predict, feed_dict={mainQN.scalarInput: [state]})[0]
next_state, next_reward, done = env.step(action)
next_reward = next_reward/600
rAll += np.sum(next_reward)
actions[i_episode,t] = action
states[i_episode,t+1] = env.state_num
plt.title('action:' + str(action) + ' time:' + str(env.time) + '\n' + 'reward:' +
np.array2string(next_reward,precision=2) +' total costs:' + str("%.2f"%rAll),**font)
plt.savefig(path1 + '/' + str(t) + '.eps')
# # plt.savefig(c_path + '/' + str(t) + '.tif')
# plt.savefig(path1 + '/' + str(t) + '.png')
def render(self):
plot_state(self.state, 7, 7)