def __init__(self, ):
super(myns3env, self).__init__()
port = 9999
simTime = 45
stepTime = 0.2
startSim = 0
seed = 3
simArgs = {
"--duration": simTime,
}
debug = True
max_env_steps = 250
self.env = ns3env.Ns3Env(port=port,
stepTime=stepTime,
startSim=startSim,
simSeed=seed,
simArgs=simArgs,
debug=debug)
self.env._max_episode_steps = max_env_steps
self.Cell_num = 6
self.max_throu = 30
self.Users = 40
self.state_dim = self.Cell_num * 4
self.action_dim = self.env.action_space.shape[0]
self.action_bound = self.env.action_space.high
self.action_space = spaces.Box(low=-1,
high=1,
shape=(self.action_dim, ),
dtype=np.float32)
self.observation_space = spaces.Box(low=0,
high=self.Users,
shape=(self.state_dim, ),
dtype=np.float32)
def env_init():
startSim = 0
port = 5555;
simTime = 20 # seconds
stepTime = 1.0 # seconds
seed = 0
simArgs = {"--simTime": simTime,
"--testArg": 123}
debug = False
env = ns3env.Ns3Env(port=port, stepTime=stepTime, startSim=startSim, simSeed=seed, simArgs=simArgs, debug=debug)
env.reset()
return env
def __init__(self, worker_name, env_count, sess, global_actor,
global_critic, max_episode_num):
threading.Thread.__init__(self)
# hyperparameters
self.GAMMA = 0.95
self.ACTOR_LEARNING_RATE = 0.0001
self.CRITIC_LEARNING_RATE = 0.001
self.ENTROPY_BETA = 0.01
self.t_MAX = 4 # n-step TD
self.max_episode_num = max_episode_num
#self.env = gym.make(env_name)
self.env = ns3env.Ns3Env(port=(port + env_count),
stepTime=stepTime,
startSim=startSim,
simSeed=seed,
simArgs=simArgs,
debug=debug)
self.worker_name = worker_name
self.sess = sess
self.global_actor = global_actor
self.global_critic = global_critic
# state variable dimension
#self.state_dim = self.env.observation_space.shape[0]
self.state_dim = state_dimension
# action dimension
#self.action_dim = self.env.action_space.shape[0]
self.action_dim = action_dimension
# action maximum boundary
#self.action_bound = int(self.env.action_space.high[0])
self.action_bound = action_max_bound
# create worker actor and critic NN
self.worker_actor = Worker_Actor(self.sess, self.state_dim,
self.action_dim, self.action_bound,
self.ACTOR_LEARNING_RATE,
self.ENTROPY_BETA, self.global_actor)
self.worker_critic = Worker_Critic(self.sess, self.state_dim,
self.action_dim,
self.CRITIC_LEARNING_RATE,
self.global_critic)
# Copy Hyperparameters from Global NN to Worker NN
self.worker_actor.model.set_weights(
self.global_actor.model.get_weights())
self.worker_critic.model.set_weights(
self.global_critic.model.get_weights())
def __init__(self, no_threads, **params):
self.params = params
self.no_threads = no_threads
self.ports = [13968+i+np.random.randint(40000) for i in range(no_threads)]
self.commands = self._craft_commands(params)
self.SCRIPT_RUNNING = False
self.envs = []
self.run()
for port in self.ports:
env = ns3env.Ns3Env(port=port, stepTime=params['envStepTime'], startSim=0, simSeed=0, simArgs=params, debug=False)
self.envs.append(env)
self.SCRIPT_RUNNING = True
def __init__(self, env_name):
# set tensorflow session
self.sess = tf.Session()
K.set_session(self.sess)
# generate learning environment
self.env_name = env_name
self.WORKERS_NUM = multiprocessing.cpu_count()
#env = gym.make(self.env_name) # change here to ns3gym
env = ns3env.Ns3Env(port=port, stepTime=stepTime, startSim=startSim, simSeed=seed, simArgs=simArgs, debug=debug)
# state dimension
#state_dim = env.observation_space.shape[0]
state_dim = state_dimension
# action dimension
#action_dim = env.action_space.shape[0]
action_dim = action_dimension
# maximum action boundary
#action_bound = env.action_space.high[0]
action_bound = action_max_bound
# generate global actor and critic nn
self.global_actor = Global_Actor(state_dim, action_dim, action_bound)
self.global_critic = Global_Critic(state_dim)
startSim = True
iterationNum = 1
port = 5555
simTime = 1000
stepTime = 0.5
seed = 12
simArgs = {
"--duration": simTime,
}
debug = False
env = ns3env.Ns3Env(port=port,
stepTime=stepTime,
startSim=startSim,
simSeed=seed,
simArgs=simArgs,
debug=debug)
env.reset()
state_space = env.observation_space.shape[0]
print(state_space)
state_trans_space = 3
action_space = 5
def get_agent(state):
socketUuid = state[0]
tcpEnvType = state[1]
tcpAgent = get_agent.tcpAgents.get(socketUuid, None)
print(tcpAgent)
elif argumentList.__len__() is 4:
if sys.argv[1] in ['test']:
# Collect from the CLI the name of the traffic scenario
scenario_name = sys.argv[2].split("=")[1]
# Collect from the CLI the number of cars that the agent was trained on
num_of_vehicles = sys.argv[3].split("=")[1]
# Load the previously trained agent parameters and start
# running the traffic simulation
# Creating the ns3 environment that will act as a link
# between our agent and the live simulation
env = ns3env.Ns3Env(port=5555,
stepTime=0.5,
startSim=0,
simSeed=12,
debug=False)
ob_space = env.observation_space
ac_space = env.action_space
print("Observation Space: ", ob_space, ob_space.dtype)
print("Action Space: ", ac_space, ac_space.dtype)
stepIdx, currIt = 0, 0
try:
# model = PPO2.load(f'rsu_agents/{scenario_name}_agents/'
# f'PPO2_ns3_online_{scenario_name}_cars={num_of_vehicles}')
#plt.savefig('learning.pdf', bbox_inches='tight')
plt.tight_layout()
plt.subplots_adjust(left = 0.16, bottom = 0.10, right = 0.92, top = 0.92, wspace = 0.20, hspace = 0.53)
plt.show()
parser = argparse.ArgumentParser(description='Start simulation script on/off')
parser.add_argument('--start',
type=int,
default=1,
help='Start ns-3 simulation script 0/1, Default: 1')
args = parser.parse_args()
startSim = bool(args.start)
print("make env")
env = ns3env.Ns3Env(port=setting.port, stepTime=setting.stepTime, startSim=startSim,
simSeed=setting.seed, simArgs=setting.simArgs, debug=setting.debug)
print("done env");
env._max_episode_steps = setting.MAX_STEPS
delay_history = []
rew_history = []
util_history = []
signal.signal(signal.SIGINT, signal_handler)
actor = ActorNetwork(state_size=setting.STATE_SIZE, action_size=setting.ACTION_SIZE, lr=setting.ACTOR_LEARNING_RATE, n_h1=setting.N_H1, n_h2=setting.N_H2, tau=setting.TAU)
critic = CriticNetwork(state_size=setting.STATE_SIZE, action_size=setting.ACTION_SIZE, lr=setting.CRITIC_LEARNING_RATE, n_h1=setting.N_H1, n_h2=setting.N_H2, tau=setting.TAU)
noise = OUProcess(setting.ACTION_SIZE)
exprep = ExpReplay(mem_size=setting.MEM_SIZE, start_mem=setting.START_MEM, state_size=[setting.STATE_SIZE], kth=-1, batch_size=setting.BATCH_SIZE)
sess = tf.Session()
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
import gym
import argparse
from ns3gym import ns3env
__author__ = "Piotr Gawlowicz"
__copyright__ = "Copyright (c) 2018, Technische Universität Berlin"
__version__ = "0.1.0"
__email__ = "*****@*****.**"
#env = gym.make('ns3-v0', port=5555)
env = ns3env.Ns3Env(port=5555)
env.reset()
ob_space = env.observation_space
ac_space = env.action_space
print("Observation space: ", ob_space, ob_space.dtype)
print("Action space: ", ac_space, ac_space.dtype)
stepIdx = 0
try:
obs = env.reset()
print("Step: ", stepIdx)
print("---obs: ", obs)
while True:
stepIdx += 1
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
import argparse
from ns3gym import ns3env
__author__ = "Piotr Gawlowicz"
__copyright__ = "Copyright (c) 2020, Technische Universität Berlin"
__version__ = "0.1.0"
__email__ = "*****@*****.**"
port = 5556
env = ns3env.Ns3Env(port=port, startSim=False)
env.reset()
ob_space = env.observation_space
ac_space = env.action_space
print("Observation space: ", ob_space, ob_space.dtype)
print("Action space: ", ac_space, ac_space.dtype)
stepIdx = 0
currIt = 0
iterationNum = 3
try:
while True:
obs = env.reset()
print("Step: ", stepIdx)
print("---obs: ", obs)
while True:
__email__ = "*****@*****.**"
parser = argparse.ArgumentParser(description='Start simulation script on/off')
parser.add_argument('--start',
type=int,
default=1,
help='Start simulation script True/False, Default: True')
args = parser.parse_args()
startSim = bool(args.start)
print("startSim: ", startSim)
stepTime = 0.5 # seconds
env = ns3env.Ns3Env(stepTime=stepTime, startSim=startSim)
env.reset()
ob_space = env.observation_space
ac_space = env.action_space
print("Observation space: ", ob_space, ob_space.dtype)
print("Action space: ", ac_space, ac_space.dtype)
stepIdx = 0
try:
while True:
stepIdx += 1
print("Step: ", stepIdx)
action = env.action_space.sample()