exit(1)
if agent_host.receivedArgument("help"):
print(agent_host.getUsage())
exit(0)
with open('world.xml', 'r') as file:
missionXML = file.read()
my_client_pool = MalmoPython.ClientPool()
my_client_pool.add(MalmoPython.ClientInfo('127.0.0.1', 10000))
EPISODES = int(config.get('DEFAULT', 'EPISODES'))
state_size = int(config.get('DEFAULT', 'STATE_SIZE'))
action_size = int(config.get('DEFAULT', 'ACTION_SIZE'))
time_multiplier = int(config.get('DEFAULT', 'TIME_MULTIPLIER'))
nn = DQNAgent(state_size, action_size)
done = False
batch_size = int(config.get('DEFAULT', 'BATCH_SIZE'))
CLEARS = 0
MAX_SUCCESS_RATE = 0
GRAPH = live_graph.Graph()
REWARDS_DICT = {}
ALL_REWARDS = []
timestep = 0
# command line arguments
try:
arg_check = sys.argv[1].lower() # using arguments from command line
if (arg_check not in ["zombie", "skeleton", "spider", "giant"]):
print("\nInvalid mob type, defaulting to 1 zombie")
mob_type = 'zombie'
import time as timer
if __name__ == "__main__":
##################################################################################
##################### Uncomment for your own ####################################
#pybulletPath = "/home/auggienanz/bullet3/data/" #Auggie
#pybulletPath = "D:/ECE 285 - Advances in Robot Manipulation/bullet3-master/data/" #Bharat
pybulletPath = 'C:/Users/Juan Camilo Castillo/Documents/bullet3/bullet3-master/data/' #Juan
outputpath = 'C:/Users/Juan Camilo Castillo/Documents/ECE 285 Robotics/save/' #Juan
#################################################################################
env = environment(pybulletPath, useGUI=True, movement_delta=0.003)
state_size = 6
action_size = 6
agent = DQNAgent(state_size, action_size)
agent.load("./run_results/JengaLearn_11.h5")
for e in range(10):
state = env.reset_random()
#print(state)
state = np.reshape(state, [1, state_size])
#timer.sleep(1)
TotalReward = 0
print('Starting Policy Rolloout from learned weights')
for time in range(300):
action = np.argmax(agent.model.predict(state)[0])
next_state, reward, done = env.step(action)
timer.sleep(0.005)
next_state = np.reshape(next_state, [1, state_size])
TotalReward = reward + TotalReward
state = next_state
ACTION_SAPCE = []
for i in BITRATE:
for j in TARGET_BUFFER:
for k in LATENCY_LIMIT:
action_apace = []
action_apace.append(i)
action_apace.append(j)
action_apace.append(k)
ACTION_SAPCE.append(action_apace)
STATE_SIZE = 250
ACTION_SIZE = len(BITRATE) * len(TARGET_BUFFER) * len(LATENCY_LIMIT)
BATCH_SIZE = 32
history_len = 50
done = False
agent = DQNAgent(STATE_SIZE, ACTION_SIZE)
def train(epoch, train_trace):
# path setting
TRAIN_TRACES = train_trace
video_size_file = './dataset/video_trace/sports/frame_trace_' # video trace path setting,
LogFile_Path = "./log/" # log file trace path setting,
# load the trace
all_cooked_time, all_cooked_bw, all_file_names = load_trace.load_trace(
TRAIN_TRACES)
# random_seed
random_seed = 2
video_count = 0
frame_time_len = 0.04
reward_all_sum = 0