if __name__ == '__main__':
parser = argparse.ArgumentParser(description='Assurance Monitoring for RL-based emergency braking system.')
parser.add_argument("-g", "--gui", help="set gui mode.", action="store_true")
parser.add_argument("-t", "--testing", help="set testing mode", action="store_true", default=False)
parser.add_argument("-cp", "--collect_perception", help="collect the data for perception training")
parser.add_argument("-ca", "--collect_detector", help="collect the data for detector training")
parser.add_argument("-p", "--perception", help="set the path of perception neural network")
parser.add_argument("-e", "--episode", help="set the number of episode", type=int, default=1)
args = parser.parse_args()
try:
os.environ["CUDA_DEVICE_ORDER"]="PCI_BUS_ID"
os.environ["CUDA_VISIBLE_DEVICES"]="0" # specify which GPU(s) to be used
collect = args_assertions(args)
env = SetupWorld(mass=1300, wheel_radius=0.04, dt=0.05, collect=collect)
agent = ddpgAgent(Testing=args.testing)
input_preprocessor = InputPreprocessor()
avf=AVF_search()
stopdist=6.0
print('Number of episodes :',args.episode)
for episode in range(args.episode):
#while stopdist>5.0:
agnt_number=2000;
numberofsamples=2000
initial_distance = np.random.normal(100, 1)
initial_speed = np.random.normal(38,11)
#initial_speed = avf.avf_predictor(numberofsamples,agnt_number)
if initial_speed <1 : initial_speed=1
friction=np.random.normal(0.7,0.15)
if friction<=0 : friction=0
parser = argparse.ArgumentParser(description='Assurance Monitoring for RL-based emergency braking system.')
parser.add_argument("-g", "--gui", help="set gui mode.", action="store_true")
parser.add_argument("-t", "--testing", help="set testing mode", action="store_true", default=False)
parser.add_argument("-cp", "--collect_perception", help="collect the data for perception training")
parser.add_argument("-ca", "--collect_detector", help="collect the data for detector training")
parser.add_argument("-p", "--perception", help="set the path of perception neural network")
parser.add_argument("-e", "--episode", help="set the number of episode", type=int, default=1)
args = parser.parse_args()
try:
collect = args_assertions(args)
carla_server = ServerManagerBinary({'CARLA_SERVER': os.environ["CARLA_SERVER"]})
carla_server.reset(port=2002)
carla_server.wait_until_ready()
env = SetupWorld(town=1, gui=args.gui, collect=collect, perception=args.perception)
agent = ddpgAgent(Testing=args.testing)
input_preprocessor = InputPreprocessor()
total_episode_count=[]
avf_model = tf.keras.models.load_model('./DATA/saved_model/my_model')
samples_per_iteration=100
container = np.load('./DATA/stdmean.npz')
for episode in range(args.episode):
#Generating data stream
random_seed=np.random.randint(1,10000)
np.random.seed(random_seed)
candidate_initial_speed =5+40*(np.random.rand(samples_per_iteration,1).reshape(-1,1))
random_seed=np.random.randint(1,10000)
np.random.seed(random_seed)
parser.add_argument("-e",
"--episode",
help="set the number of episode",
type=int,
default=1)
args = parser.parse_args()
try:
collect = args_assertions(args)
carla_server = ServerManagerBinary(
{'CARLA_SERVER': os.environ["CARLA_SERVER"]})
carla_server.reset()
carla_server.wait_until_ready()
env = SetupWorld(town=1,
gui=args.gui,
collect=collect,
perception=args.perception)
agent = ddpgAgent(Testing=args.testing)
input_preprocessor = InputPreprocessor()
for episode in range(args.episode):
initial_distance = np.random.normal(100, 1)
initial_speed = np.random.uniform(5, 45)
#initial_speed =45
s = env.reset(initial_distance, initial_speed)
print(
"Episode {} is started, target distance: {}, target speed: {}, initial distance: {}, initial speed: {}"
.format(episode, initial_distance, initial_speed, s[0], s[1]))
s = input_preprocessor(s)
epsilon = 1.0 - (episode + 1) / (args.episode)
while True:
a = agent.getAction(s, epsilon)
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='Assurance Monitoring for RL-based emergency braking system.')
parser.add_argument("-g", "--gui", help="set gui mode.", action="store_true")
parser.add_argument("-t", "--testing", help="set testing mode", action="store_true", default=False)
parser.add_argument("-cp", "--collect_perception", help="collect the data for perception training")
parser.add_argument("-ca", "--collect_detector", help="collect the data for detector training")
parser.add_argument("-p", "--perception", help="set the path of perception neural network")
parser.add_argument("-e", "--episode", help="set the number of episode", type=int, default=1)
args = parser.parse_args()
try:
os.environ["CUDA_DEVICE_ORDER"]="PCI_BUS_ID"
os.environ["CUDA_VISIBLE_DEVICES"]="0" # specify which GPU(s) to be used
collect = args_assertions(args)
env = SetupWorld(mass=1300, wheel_radius=0.04, dt=0.05, collect=collect)
agent = ddpgAgent(Testing=args.testing)
input_preprocessor = InputPreprocessor()
agnt_number=15000; # select agent
pr_avf=PR_AVF(agnt_number) # Priority replay avf class declaration
avf=AVF_search()
crashed = pd.read_csv('./DATA/crashed_for_train.csv')
crashed=crashed[["Kick_Speed","friction_patch","loc_patch","size_patch"]].values
print(crashed)
gmm=gmm_trainer(crashed)
number_of_retraining=0 # avf search class declaration
with open('gmm_5D', 'rb') as f:
model = pickle.load(f)
failure_collection=[];
help="set the path of perception neural network")
parser.add_argument("-e",
"--episode",
help="set the number of episode",
type=int,
default=1)
args = parser.parse_args()
try:
os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
os.environ[
"CUDA_VISIBLE_DEVICES"] = "0" # specify which GPU(s) to be used
collect = args_assertions(args)
env = SetupWorld(mass=1300,
wheel_radius=0.04,
dt=0.05,
collect=collect)
agent = ddpgAgent(Testing=args.testing)
input_preprocessor = InputPreprocessor()
rcf = rcf(100)
print('Number of episodes :', args.episode)
plot = Liveplot()
cnt = 0
coDisp = []
#rcf.trainer()
for episode in range(args.episode):
#stopdist=1.0;
#Comment below 3 lines while training , its only for running set of testing
#print('***********************************************************************************************************')
#print('******** Launching failure search test ID:',episode+1)
#np.random.seed()