currIt = 0
rd = []
plt.ion()
try:
while True:
print("Start iteration: ", currIt)
obs = env.reset()
print("Step: ", stepIdx)
print("---obs:", obs)
while True:
reward = 0
matrixOfChanAlloc = np.zeros((nOfenb, nOfchannel))
stepIdx += 1
if stepIdx % 100 == 0:
PG.ep = PG.ep * 0.7
ax.append(stepIdx)
print("stepIdx: ", stepIdx)
# ax.append(stepIdx)
# ---------------------------------------------------------------------------------------
observation = [] #环境的观测值，状态observation
for j in range((int)(len(obs) / 4)):
#状态
observation.append([
obs[4 * j], obs[4 * j + 1], obs[4 * j + 2], obs[4 * j + 3]
])
action_list = []
print("obs: ", obs)
if (len(observation) == 0):
observation_step = [0, 0, 0, 0]
ss = observation[k].copy()
plt.ion()
try:
while True:
print("Start iteration: ", currIt)
obs = env.reset()
print("Step: ", stepIdx)
print("---obs:", obs)
flag = False
while True:
reward = 0
matrixOfChanAlloc = np.zeros((nOfenb, nOfchannel))
stepIdx += 1
if stepIdx % 100 == 0:
PG.ep = PG.ep * 0.95
ax.append(stepIdx)
print("stepIdx: ", stepIdx)
print("obs: ", obs)
observation = [] #环境的观测值，状态observation
observation, numue = getObservation(observation,
obs) #将ns3的观测值转为gym可用的形式
action_list = [] #存储动作的list
if numue == 0: #若有效请求数为0，则返回一个空动作
addaction(0, 0, 0, action_list)
action_tuple = listTotuple(action_list)
obs, reward_step, done, info = env.step(
action_tuple) #获取这一eposide的奖励