def get(self):
self.driver.get(self.dic_url[self.sid])
if self.sid == 'SWGS':
try:
self.driver.switch_to_alert()
alert = WebDriverWait(self.driver, 1).until(EC.alert_is_present())
alert.accept()
except NoAlertPresentException:
pass
finally:
while True:
try:
WebDriverWait(self.driver, 0).until(EC.element_to_be_clickable((By.ID, 'bodyBlock')))
except:
break
try:
WebDriverWait(self.driver, 60).until(EC.presence_of_element_located((By.ID, 'mdi01_subWindow0_iframe')))
finally:
print('Page is ready!')
self.driver.execute_script('''top.document.title = "(FOR AUTO TEST TOOL)"''')
# Recorder 선언
self.recorder = Recorder(self.driver)
self.recorder.addEventListener()
def do_client(idx, args):
dataset = load_dataset('wmt14ende', splits=('test'))
headers = {"Content-type": "application/json"}
url = "http://127.0.0.1:9292/transformer/prediction"
batch = []
sample = 0
f = open(args.output_file, "w")
if args.profile:
recorder = Recorder(args.infer_batch_size, args.model_name)
recorder.tic()
for sequence in dataset:
sample += 1
batch.append(sequence[args.src_lang])
if len(batch) < args.infer_batch_size and sample != len(dataset):
continue
data = {"feed": [{"src_word": batch}], "fetch": ["finished_sequence"]}
r = requests.post(url=url, headers=headers, data=json.dumps(data))
if r is not None:
print("Status: ", r)
if args.profile:
recorder.toc(samples=len(batch))
else:
for seq in r.json()["result"]["finished_sequence"]:
f.write(seq[0] + "\n")
batch = []
if args.profile:
recorder.tic()
f.close()
if args.profile:
recorder.report()
return [[recorder.infer_time]]
def __init__(self, task, train_exp_dir, ite, logdir=None):
self.task = task
self.policy = LoadPolicy('../utils/models/{}/{}'.format(task, train_exp_dir), ite)
self.env = CrossroadEnd2end(training_task=self.task, mode='testing')
self.model = EnvironmentModel(self.task, mode='selecting')
self.recorder = Recorder()
self.episode_counter = -1
self.step_counter = -1
self.obs = None
self.stg = MultiPathGenerator()
self.step_timer = TimerStat()
self.ss_timer = TimerStat()
self.logdir = logdir
if self.logdir is not None:
config = dict(task=task, train_exp_dir=train_exp_dir, ite=ite)
with open(self.logdir + '/config.json', 'w', encoding='utf-8') as f:
json.dump(config, f, ensure_ascii=False, indent=4)
self.fig = plt.figure(figsize=(8, 8))
plt.ion()
self.hist_posi = []
self.old_index = 0
self.path_list = self.stg.generate_path(self.task)
# ------------------build graph for tf.function in advance-----------------------
for i in range(3):
obs = self.env.reset()
obs = tf.convert_to_tensor(obs[np.newaxis, :], dtype=tf.float32)
self.is_safe(obs, i)
obs = self.env.reset()
obs_with_specific_shape = np.tile(obs, (3, 1))
self.policy.run_batch(obs_with_specific_shape)
self.policy.obj_value_batch(obs_with_specific_shape)
# ------------------build graph for tf.function in advance-----------------------
self.reset()
def generate_recorder(self, logger2file, graph):
"""
create model/log/data dictionary and define writer to record training data.
"""
self.check_or_create(self.cp_dir, 'checkpoints')
self.check_or_create(self.log_dir, 'logs(summaries)')
self.check_or_create(self.excel_dir, 'excel')
self.recorder = Recorder(log_dir=self.log_dir,
excel_dir=self.excel_dir,
logger2file=logger2file,
graph=graph)
def start():
global recorder, logger, ConfigFile
if recorder is None or recorder.is_stopped():
logger.log('Creating new PCRC recorder')
try:
recorder = Recorder(ConfigFile, TranslationFolder)
except YggdrasilError as e:
logger.error(e)
return
ret = recorder.start()
logger.log('Recorder started, success = {}'.format(ret))
else:
logger.warn('Recorder is running, ignore')
def run(self, category, im_list, output_fn="tmp.txt"):
# Load checkpoint
checkpoint_dir = "{}/{}/{}".format(self.main_dir, category, self.lr)
checkpoint, _ = get_latest_checkpoint(checkpoint_dir)
disc = Discriminator(lr=self.lr, checkpoint=checkpoint)
output_dir = join(self.main_dir, "predictions")
output_path = join(output_dir, output_fn)
recorder = Recorder(output_path, restart=False)
for im in im_list:
if recorder.contains(im):
print im, recorder.get(im), "Done."
continue
img, _ = self.datasource.get_image(im)
gt, _ = self.datasource.get_ground_truth(im, one_hot=True)
ap, _ = self.datasource.get_all_prob(im)
pr_prob = disc.predict(img, ap, args.category)
gt_prob = disc.predict(img, gt, args.category)
print "{} {} {}".format(im, pr_prob, gt_prob)
recorder.save(im, [pr_prob, gt_prob])
recorder.write()
return output_path
def train(
_net, _train_loader, _optimizer, _criterion, _device = 'cpu',
_recorder: Recorder = None,
_weight_quantization_error_collection = None, _input_quantization_error_collection = None,
_weight_bit_allocation_collection = None, _input_bit_allocation_collection = None
):
_net.train()
_train_loss = 0
_correct = 0
_total = 0
for batch_idx, (inputs, targets) in enumerate(_train_loader):
inputs, targets = inputs.to(_device), targets.to(_device)
_optimizer.zero_grad()
outputs = _net(inputs)
losses = _criterion(outputs, targets)
losses.backward()
_optimizer.step()
_train_loss += losses.data.item()
_, predicted = torch.max(outputs.data, 1)
_total += targets.size(0)
_correct += predicted.eq(targets.data).cpu().sum().item()
progress_bar(
batch_idx, len(_train_loader),
'Loss: %.3f | Acc: %.3f%% (%d/%d)' % (_train_loss / (batch_idx + 1), 100. * _correct / _total, _correct, _total)
)
if _recorder is not None:
_recorder.update(loss=losses.data.item(), acc=[_correct / _total], batch_size=inputs.size(0), is_train=True)
if _weight_quantization_error_collection and _input_quantization_error_collection is not None:
for name, layer in _net.quantized_layer_collections.items():
_weight_quantization_error = torch.abs(layer.quantized_weight - layer.pre_quantized_weight).mean().item()
_input_quantization_error = torch.abs(layer.quantized_input - layer.pre_quantized_input).mean().item()
_weight_quantization_error_collection[name].write('%.8e\n' % _weight_quantization_error)
_input_quantization_error_collection[name].write('%.8e\n' % _input_quantization_error)
if _weight_bit_allocation_collection and _input_bit_allocation_collection is not None:
for name, layer in _net.quantized_layer_collections.items():
_weight_bit_allocation_collection[name].write('%.2f\n' % (torch.abs(layer.quantized_weight_bit).mean().item()))
_input_bit_allocation_collection[name].write('%.2f\n' % (torch.abs(layer.quantized_input_bit).mean().item()))
return _train_loss / (len(_train_loader)), _correct / _total
def __init__(self, task):
self.task = task
if self.task == 'left':
self.policy = LoadPolicy('G:\\env_build\\utils\\models\\left', 100000)
elif self.task == 'right':
self.policy = LoadPolicy('G:\\env_build\\utils\\models\\right', 145000)
elif self.task == 'straight':
self.policy = LoadPolicy('G:\\env_build\\utils\\models\\straight', 95000)
self.horizon = 25
self.num_future_data = 0
self.env = CrossroadEnd2end(training_task=self.task, num_future_data=self.num_future_data)
self.model = EnvironmentModel(self.task)
self.obs = self.env.reset()
self.stg = StaticTrajectoryGenerator_origin(mode='static_traj')
self.data2plot = []
self.mpc_cal_timer = TimerStat()
self.adp_cal_timer = TimerStat()
self.recorder = Recorder()
def __init__(self, task, logdir=None):
self.task = task
if self.task == 'left':
self.policy = LoadPolicy('../utils/models/left', 100000)
elif self.task == 'right':
self.policy = LoadPolicy('../utils/models/right', 145000)
elif self.task == 'straight':
self.policy = LoadPolicy('../utils/models/straight', 95000)
self.env = CrossroadEnd2end(training_task=self.task)
self.model = EnvironmentModel(self.task, mode='selecting')
self.recorder = Recorder()
self.episode_counter = -1
self.step_counter = -1
self.obs = None
self.stg = None
self.step_timer = TimerStat()
self.ss_timer = TimerStat()
self.logdir = logdir
self.fig = plt.figure(figsize=(8, 8))
plt.ion()
self.hist_posi = []
self.reset()
def __init__(self, task, train_exp_dir, ite, logdir=None):
self.task = task
self.policy = LoadPolicy(
'../utils/models/{}/{}'.format(task, train_exp_dir), ite)
self.env = CrossroadEnd2end(training_task=self.task)
self.model = EnvironmentModel(self.task, mode='selecting')
self.recorder = Recorder()
self.episode_counter = -1
self.step_counter = -1
self.obs = None
self.stg = None
self.step_timer = TimerStat()
self.ss_timer = TimerStat()
self.logdir = logdir
if self.logdir is not None:
config = dict(task=task, train_exp_dir=train_exp_dir, ite=ite)
with open(self.logdir + '/config.json', 'w',
encoding='utf-8') as f:
json.dump(config, f, ensure_ascii=False, indent=4)
self.fig = plt.figure(figsize=(8, 8))
plt.ion()
self.hist_posi = []
self.reset()
def create_predictor(cls,
args,
config=None,
profile=False,
model_name=None):
if config is None:
config = inference.Config(
os.path.join(args.inference_model_dir, "transformer.pdmodel"),
os.path.join(args.inference_model_dir,
"transformer.pdiparams"))
if args.device == "gpu":
config.enable_use_gpu(100, 0)
elif args.device == "xpu":
config.enable_xpu(100)
else:
# CPU
config.disable_gpu()
if args.use_mkl:
config.enable_mkldnn()
config.set_cpu_math_library_num_threads(args.threads)
# Use ZeroCopy.
config.switch_use_feed_fetch_ops(False)
if profile:
recorder = Recorder(config, args.infer_batch_size, model_name)
else:
recorder = None
predictor = inference.create_predictor(config)
input_handles = [
predictor.get_input_handle(name)
for name in predictor.get_input_names()
]
output_handles = [
predictor.get_input_handle(name)
for name in predictor.get_output_names()
]
return cls(predictor, input_handles, output_handles, recorder)
def setup_recorder(params):
recorder = Recorder()
# This is for early stopping, currectly I did not use it
recorder.bad_counter = 0 # start from 0
recorder.estop = False
recorder.lidx = -1 # local data index
recorder.step = 0 # global step, start from 0
recorder.epoch = 1 # epoch number, start from 1
recorder.history_scores = []
recorder.valid_script_scores = []
# trying to load saved recorder
record_path = os.path.join(params.output_dir, "record.json")
record_path = os.path.abspath(record_path)
if tf.gfile.Exists(record_path):
recorder.load_from_json(record_path)
params.add_hparam('recorder', recorder)
return params
if use_cuda:
net.cuda()
# net = torch.nn.DataParallel(net, device_ids=range(torch.cuda.device_count()))
cudnn.benchmark = True
# ---
# Begin Training
# ---
ascent_count = 0
min_train_loss = 1e9
criterion = nn.CrossEntropyLoss()
max_training_epoch = args.max_epoch
# Initialize recorder for general training
recorder = Recorder(SummaryPath=save_root)
recorder.write_arguments([args])
# Initialize recorder for threshold
weight_quantization_error_recorder_collection = {}
input_quantization_error_recorder_collection = {}
weight_bit_allocation_collection = {}
input_bit_allocation_collection = {}
for name, layer in net.quantized_layer_collections.items():
if not os.path.exists('%s/%s' % (save_root, name)):
os.makedirs('%s/%s' % (save_root, name))
weight_quantization_error_recorder_collection[name] = open('%s/%s/weight_quantization_error.txt' % (save_root, name), 'a+')
input_quantization_error_recorder_collection[name] = open('%s/%s/input_quantization_error.txt' % (save_root, name), 'a+')
weight_bit_allocation_collection[name] = open('%s/%s/weight_bit_allocation.txt' % (save_root, name), 'a+')
input_bit_allocation_collection[name] = open('%s/%s/input_bit_allocation.txt' % (save_root, name), 'a+')
for epoch in range(start_epoch, start_epoch + max_training_epoch):
import wave, audioop, time
from aip import AipSpeech
from utils.recorder import Recorder
from utils import gmm_vad
import threading
UPLOAD_PATH = os.path.join(basedir, 'static/uploads/')
VOICES_PATH = os.path.join(UPLOAD_PATH, 'voices/')
TEMP_PATH = os.path.join(UPLOAD_PATH, 'temps/')
ALGORITHM_PATH = os.path.join(UPLOAD_PATH, 'algorithms')
MODELS_PATH = os.path.join(UPLOAD_PATH, 'models')
sys.path.append(ALGORITHM_PATH)
sys.path.append(MODELS_PATH)
started_list = []
rec = Recorder()
class TokenResource(Resource):
# decorators = [auth.login_required]
def get(self):
token = g.user.generate_auth_token()
return jsonify({'token': token.decode('ascii')})
class LoginResource(Resource):
def post(self):
username = request.json['username']
password = request.json['password']
result = {}
if (self.verify_auth(username, password)):
from utils.recorder import Recorder
from sorting.merge_sort import merge_sort
from sorting.merge_sort_iterative import merge_sort_iter
from sorting.quick_sort import quick_sort
from sorting.quick_sort_3way import quick_sort_3way
from sorting.selection_sort import selection_sort
from sorting.insertion_sort import insert_sort_optimized
from sorting.insertion_sort import insert_sort
import numpy as np
from time import time
from threading import Thread
if __name__ == '__main__':
s = time()
recorder = Recorder()
start, end = 2, 6
for x in range(start, end):
data = list(np.random.rand(np.power(10, x)))
p1 = Thread(target=recorder.execute, args=(selection_sort, data))
p2 = Thread(target=recorder.execute, args=(insert_sort, data))
p3 = Thread(target=recorder.execute,
args=(insert_sort_optimized, data))
p4 = Thread(target=recorder.execute,
args=(merge_sort, data, [0, len(data) - 1]))
p5 = Thread(target=recorder.execute, args=(merge_sort_iter, data))
p6 = Thread(target=recorder.execute,
args=(quick_sort, data, [0, len(data) - 1]))
# Initial Recorder #
####################
if args.exp_spec is not '':
SummaryPath += ('-' + args.exp_spec)
print('Save to %s' % SummaryPath)
if os.path.exists(SummaryPath):
print('Record exist, remove')
input()
shutil.rmtree(SummaryPath)
os.makedirs(SummaryPath)
else:
os.makedirs(SummaryPath)
recorder = Recorder(SummaryPath=SummaryPath, dataset_name=dataset_name)
##################
# Begin Training #
##################
meta_grad_dict = dict()
for epoch in range(MAX_EPOCH):
if recorder.stop: break
print('\nEpoch: %d, lr: %e' % (epoch, optimizee.param_groups[0]['lr']))
net.train()
end = time.time()
recorder.reset_performance()
def plot_data(logdir, i):
recorder = Recorder()
recorder.load(logdir)
recorder.plot_ith_episode_curves(i)
class Base:
def __init__(self, *args, **kwargs):
'''
inputs:
a_dim: action spaces
is_continuous: action type, refer to whether this control problem is continuous(True) or discrete(False)
base_dir: the directory that store data, like model, logs, and other data
'''
super().__init__()
base_dir = kwargs.get('base_dir')
tf_dtype = str(kwargs.get('tf_dtype'))
tf.random.set_seed(int(kwargs.get('seed', 0)))
self.device = get_device()
self.logger2file = bool(kwargs.get('logger2file', False))
tf.keras.backend.set_floatx(tf_dtype)
self.cp_dir, self.log_dir, self.excel_dir = [
os.path.join(base_dir, i) for i in ['model', 'log', 'excel']
]
self.global_step = tf.Variable(
0, name="global_step", trainable=False, dtype=tf.int64
) # in TF 2.x must be tf.int64, because function set_step need args to be tf.int64.
self.cast = self._cast(dtype=tf_dtype)
def _cast(self, dtype='float32'):
'''
TODO: Annotation
'''
if dtype == 'float32':
func = cast2float32
self._data_type = tf.float32
elif dtype == 'float64':
self._data_type = tf.float64
func = cast2float64
else:
raise Exception('Cast to this type has not been implemented.')
def inner(*args, **kwargs):
with tf.device(self.device):
return func(*args, **kwargs)
return inner
def data_convert(self, data):
'''
TODO: Annotation
'''
return tf.convert_to_tensor(data, dtype=self._data_type)
def get_init_train_step(self):
"""
get the initial training step. use for continue train from last training step.
"""
if os.path.exists(os.path.join(self.cp_dir, 'checkpoint')):
return int(tf.train.latest_checkpoint(self.cp_dir).split('-')[-1])
else:
return 0
def get_init_training_info(self):
'''
TODO: Annotation
'''
return self.recorder.get_training_info()
def generate_recorder(self, kwargs):
"""
create model/log/data dictionary and define writer to record training data.
"""
self.check_or_create(self.cp_dir, 'checkpoints')
self.check_or_create(self.log_dir, 'logs(summaries)')
self.check_or_create(self.excel_dir, 'excel')
self.recorder = Recorder(kwargs,
cp_dir=self.cp_dir,
log_dir=self.log_dir,
excel_dir=self.excel_dir,
logger2file=self.logger2file)
def init_or_restore(self, base_dir):
"""
check whether chekpoint and model be within cp_dir, if in it, restore otherwise initialize randomly.
"""
cp_dir = os.path.join(base_dir, 'model')
if os.path.exists(os.path.join(cp_dir, 'checkpoint')):
try:
self.recorder.checkpoint.restore(
tf.train.latest_checkpoint(
cp_dir)).expect_partial() # 从指定路径导入模型
except:
self.recorder.logger.error(
f'restore model from {cp_dir} FAILED.')
raise Exception(f'restore model from {cp_dir} FAILED.')
else:
self.recorder.logger.info(
f'restore model from {cp_dir} SUCCUESS.')
else:
self.recorder.checkpoint.restore(
self.recorder.saver.latest_checkpoint).expect_partial(
) # 从本模型目录载入模型，断点续训
self.recorder.logger.info(
f'restore model from {self.recorder.saver.latest_checkpoint} SUCCUESS.'
)
self.recorder.logger.info('initialize model SUCCUESS.')
def save_checkpoint(self, **kwargs):
"""
save the training model
"""
train_step = int(kwargs.get('train_step', 0))
self.recorder.saver.save(checkpoint_number=train_step)
logger.info(f'Save checkpoint success. Training step: {train_step}')
self.recorder.write_training_info(kwargs)
def writer_summary(self, global_step, **kargs):
"""
record the data used to show in the tensorboard
"""
self.recorder.writer.set_as_default()
tf.summary.experimental.set_step(global_step)
for i in [{
'tag': 'AGENT/' + key,
'value': kargs[key]
} for key in kargs]:
tf.summary.scalar(i['tag'], i['value'])
self.recorder.writer.flush()
def write_training_summaries(self, global_step, summaries: dict):
'''
write tf summaries showing in tensorboard.
'''
self.recorder.writer.set_as_default()
tf.summary.experimental.set_step(global_step)
for key, value in summaries.items():
tf.summary.scalar(key, value)
self.recorder.writer.flush()
def check_or_create(self, dicpath, name=''):
"""
check dictionary whether existing, if not then create it.
"""
if not os.path.exists(dicpath):
os.makedirs(dicpath)
logger.info(''.join([f'create {name} directionary :', dicpath]))
def save_weights(self, path):
"""
save trained weights
:return: None
"""
# self.net.save_weights(os.path.join(path, 'net.ckpt'))
pass
def load_weights(self, path):
"""
load trained weights
:return: None
"""
# self.net.load_weights(os.path.join(path, 'net.ckpt'))
pass
def close(self):
"""
end training, and export the training model
"""
pass
def get_global_step(self):
"""
get the current training step.
"""
return self.global_step
def set_global_step(self, num):
"""
set the start training step.
"""
self.global_step.assign(num)
def show_logo(self):
raise NotImplementedError
def main():
if sys.platform.startswith('win'):
# Add the _win_handler function to the windows console's handler function list
win32api.SetConsoleCtrlHandler(_win_handler, True)
if os.path.exists(
os.path.join(config_file.config['config_file'], 'config.yaml')):
config = sth.load_config(config_file.config['config_file'])
else:
config = config_file.config
print(f'load config from config.')
hyper_config = config['hyper parameters']
train_config = config['train config']
record_config = config['record config']
basic_dir = record_config['basic_dir']
last_name = record_config['project_name'] + '/' \
+ record_config['remark'] \
+ record_config['run_id']
cp_dir = record_config['checkpoint_basic_dir'] + last_name
cp_file = cp_dir + '/rb'
log_dir = record_config['log_basic_dir'] + last_name
excel_dir = record_config['excel_basic_dir'] + last_name
config_dir = record_config['config_basic_dir'] + last_name
sth.check_or_create(basic_dir, 'basic')
sth.check_or_create(cp_dir, 'checkpoints')
sth.check_or_create(log_dir, 'logs(summaries)')
sth.check_or_create(excel_dir, 'excel')
sth.check_or_create(config_dir, 'config')
logger = create_logger(
name='logger',
console_level=logging.INFO,
console_format='%(levelname)s : %(message)s',
logger2file=record_config['logger2file'],
file_name=log_dir + '\log.txt',
file_level=logging.WARNING,
file_format=
'%(lineno)d - %(asctime)s - %(module)s - %(funcName)s - %(levelname)s - %(message)s'
)
if train_config['train']:
sth.save_config(config_dir, config)
if train_config['unity_mode']:
env = UnityEnvironment()
else:
env = UnityEnvironment(
file_name=train_config['unity_file'],
no_graphics=True if train_config['train'] else False,
base_port=train_config['port'])
brain_name = env.external_brain_names[0]
brain = env.brains[brain_name]
# set the memory use proportion of GPU
tf_config = tf.ConfigProto()
tf_config.gpu_options.allow_growth = True
# tf_config.gpu_options.per_process_gpu_memory_fraction = 0.5
tf.reset_default_graph()
graph = tf.Graph()
with graph.as_default() as g:
with tf.Session(graph=g, config=tf_config) as sess:
logger.info('Algorithm: {0}'.format(
train_config['algorithm'].name))
if train_config['algorithm'] == config_file.algorithms.ppo_sep_ac:
from ppo.ppo_base import PPO_SEP
model = PPO_SEP(sess=sess,
s_dim=brain.vector_observation_space_size,
a_counts=brain.vector_action_space_size[0],
hyper_config=hyper_config)
logger.info('PPO_SEP initialize success.')
elif train_config['algorithm'] == config_file.algorithms.ppo_com:
from ppo.ppo_base import PPO_COM
model = PPO_COM(sess=sess,
s_dim=brain.vector_observation_space_size,
a_counts=brain.vector_action_space_size[0],
hyper_config=hyper_config)
logger.info('PPO_COM initialize success.')
elif train_config['algorithm'] == config_file.algorithms.sac:
from sac.sac import SAC
model = SAC(sess=sess,
s_dim=brain.vector_observation_space_size,
a_counts=brain.vector_action_space_size[0],
hyper_config=hyper_config)
logger.info('SAC initialize success.')
elif train_config['algorithm'] == config_file.algorithms.sac_no_v:
from sac.sac_no_v import SAC_NO_V
model = SAC_NO_V(sess=sess,
s_dim=brain.vector_observation_space_size,
a_counts=brain.vector_action_space_size[0],
hyper_config=hyper_config)
logger.info('SAC_NO_V initialize success.')
elif train_config['algorithm'] == config_file.algorithms.ddpg:
from ddpg.ddpg import DDPG
model = DDPG(sess=sess,
s_dim=brain.vector_observation_space_size,
a_counts=brain.vector_action_space_size[0],
hyper_config=hyper_config)
logger.info('DDPG initialize success.')
elif train_config['algorithm'] == config_file.algorithms.td3:
from td3.td3 import TD3
model = TD3(sess=sess,
s_dim=brain.vector_observation_space_size,
a_counts=brain.vector_action_space_size[0],
hyper_config=hyper_config)
logger.info('TD3 initialize success.')
recorder = Recorder(log_dir,
excel_dir,
record_config,
logger,
max_to_keep=5,
pad_step_number=True,
graph=g)
episode = init_or_restore(cp_dir, sess, recorder, cp_file)
try:
if train_config['train']:
train_OnPolicy(
sess=sess,
env=env,
brain_name=brain_name,
begin_episode=episode,
model=model,
recorder=recorder,
cp_file=cp_file,
hyper_config=hyper_config,
train_config=train_config) if not train_config[
'use_replay_buffer'] else train_OffPolicy(
sess=sess,
env=env,
brain_name=brain_name,
begin_episode=episode,
model=model,
recorder=recorder,
cp_file=cp_file,
hyper_config=hyper_config,
train_config=train_config)
tf.train.write_graph(g,
cp_dir,
'raw_graph_def.pb',
as_text=False)
export_model(cp_dir, g)
else:
inference(env, brain_name, model, train_config)
except Exception as e:
logger.error(e)
finally:
env.close()
recorder.close()
sys.exit()
def __init__(self, task_name, task_type = 'prune', optimizer_type = 'adam',
save_root = None, SummaryPath = None, use_cuda = True, **kwargs):
self.task_name = task_name
self.task_type = task_type # prune, soft-quantize
self.model_name, self.dataset_name = task_name.split('-')
self.ratio = 'sample' if self.dataset_name in ['CIFARS'] else -1
#######
# Net #
#######
if task_type == 'prune':
if self.model_name == 'ResNet20':
if self.dataset_name in ['CIFAR10', 'CIFARS']:
self.net = resnet20_cifar()
elif self.dataset_name == 'STL10':
self.net = resnet20_stl()
else:
raise NotImplementedError
elif self.model_name == 'ResNet32':
if self.dataset_name in ['CIFAR10', 'CIFARS']:
self.net = resnet32_cifar()
elif self.dataset_name == 'STL10':
self.net = resnet32_stl()
else:
raise NotImplementedError
elif self.model_name == 'ResNet56':
if self.dataset_name in ['CIFAR10', 'CIFARS']:
self.net = resnet56_cifar()
elif self.dataset_name == 'CIFAR100':
self.net = resnet56_cifar(num_classes=100)
elif self.dataset_name == 'STL10':
self.net = resnet56_stl()
else:
raise NotImplementedError
elif self.model_name == 'ResNet18':
if self.dataset_name == 'ImageNet':
self.net = resnet18()
else:
raise NotImplementedError
elif self.model_name == 'vgg11':
self.net = vgg11() if self.dataset_name == 'CIFAR10' else vgg11_stl10()
else:
print(self.model_name, self.dataset_name)
raise NotImplementedError
elif task_type == 'soft-quantize':
if self.model_name == 'ResNet20':
if self.dataset_name in ['CIFAR10', 'CIFARS']:
self.net = soft_quantized_resnet20_cifar()
elif self.dataset_name in ['STL10']:
self.net = soft_quantized_resnet20_stl()
else:
raise NotImplementedError
else:
raise ('Task type not defined.')
self.meta_opt_flag = True # True for enabling meta leraning
##############
# Meta Prune #
##############
self.mask_dict = dict()
self.meta_grad_dict = dict()
self.meta_hidden_state_dict = dict()
######################
# Meta Soft Quantize #
######################
self.quantized = 0 # Quantized type
self.alpha_dict = dict()
self.alpha_hidden_dict = dict()
self.sq_rate = 0
self.s_rate = 0
self.q_rate = 0
##########
# Record #
##########
self.dataset_type = 'large' if self.dataset_name in ['ImageNet'] else 'small'
self.SummaryPath = SummaryPath
self.save_root = save_root
self.recorder = Recorder(self.SummaryPath, self.dataset_name, self.task_name)
####################
# Load Pre-trained #
####################
self.pretrain_path = '%s/%s-pretrain.pth' %(self.save_root, self.task_name)
self.net.load_state_dict(torch.load(self.pretrain_path))
print('Load pre-trained model from %s' %self.pretrain_path)
if use_cuda:
self.net.cuda()
# Optimizer for this task
if optimizer_type in ['Adam', 'adam']:
self.optimizer = Adam(self.net.parameters(), lr=1e-3)
else:
self.optimizer = SGD(self.net.parameters())
if self.dataset_name == 'ImageNet':
try:
self.train_loader = get_lmdb_imagenet('train', 128)
self.test_loader = get_lmdb_imagenet('test', 100)
except:
self.train_loader = get_dataloader(self.dataset_name, 'train', 128)
self.test_loader = get_dataloader(self.dataset_name, 'test', 100)
else:
self.train_loader = get_dataloader(self.dataset_name, 'train', 128, ratio=self.ratio)
self.test_loader = get_dataloader(self.dataset_name, 'test', 128)
self.iter_train_loader = yielder(self.train_loader)
class Base(object):
_version_number_ = 2
def __init__(self, a_dim_or_list, action_type, base_dir):
self.graph = tf.Graph()
gpu_options = tf.GPUOptions(allow_growth=True)
self.sess = tf.Session(config=tf.ConfigProto(gpu_options=gpu_options),
graph=self.graph)
self.cp_dir, self.log_dir, self.excel_dir = [
os.path.join(base_dir, i) for i in ['model', 'log', 'excel']
]
self.action_type = action_type
self.a_counts = int(np.array(a_dim_or_list).prod())
self.possible_output_nodes = [
'action', 'version_number', 'is_continuous_control',
'action_output_shape', 'memory_size'
]
with self.graph.as_default():
tf.set_random_seed(-1) # variables initialization consistent.
tf.Variable(1 if action_type == 'continuous' else 0,
name='is_continuous_control',
trainable=False,
dtype=tf.int32) # continuous 1 discrete 0
tf.Variable(self.a_counts,
name="action_output_shape",
trainable=False,
dtype=tf.int32)
tf.Variable(self._version_number_,
name='version_number',
trainable=False,
dtype=tf.int32)
tf.Variable(0, name="memory_size", trainable=False, dtype=tf.int32)
self.episode = tf.Variable(tf.constant(0))
self.global_step = tf.Variable(0,
name="global_step",
trainable=False,
dtype=tf.int64)
def get_init_episode(self):
"""
get the initial training step. use for continue train from last training step.
"""
if os.path.exists(os.path.join(self.cp_dir, 'checkpoint')):
return int(tf.train.latest_checkpoint(self.cp_dir).split('-')[-1])
else:
return 0
def generate_recorder(self, logger2file, graph):
"""
create model/log/data dictionary and define writer to record training data.
"""
self.check_or_create(self.cp_dir, 'checkpoints')
self.check_or_create(self.log_dir, 'logs(summaries)')
self.check_or_create(self.excel_dir, 'excel')
self.recorder = Recorder(log_dir=self.log_dir,
excel_dir=self.excel_dir,
logger2file=logger2file,
graph=graph)
def init_or_restore(self, base_dir):
"""
check whether chekpoint and model be within cp_dir, if in it, restore otherwise initialize randomly.
"""
cp_dir = os.path.join(base_dir, 'model')
with self.graph.as_default():
if os.path.exists(os.path.join(cp_dir, 'checkpoint')):
try:
self.recorder.saver.restore(
self.sess, tf.train.latest_checkpoint(cp_dir))
except:
self.recorder.logger.error(
'restore model from checkpoint FAILED.')
else:
self.recorder.logger.info(
'restore model from checkpoint SUCCUESS.')
else:
self.sess.run(tf.global_variables_initializer())
self.recorder.logger.info('initialize model SUCCUESS.')
def save_checkpoint(self, global_step):
"""
save the training model
"""
self.recorder.saver.save(self.sess,
os.path.join(self.cp_dir, 'rb'),
global_step=global_step,
write_meta_graph=False)
def writer_summary(self, global_step, **kargs):
"""
record the data used to show in the tensorboard
"""
self.recorder.writer_summary(x=global_step,
ys=[{
'tag': 'MAIN/' + key,
'value': kargs[key]
} for key in kargs])
def _process_graph(self):
"""
Gets the list of the output nodes present in the graph for inference
:return: list of node names
"""
all_nodes = [x.name for x in self.graph.as_graph_def().node]
nodes = [x for x in all_nodes if x in self.possible_output_nodes]
return nodes
def export_model(self):
"""
Exports latest saved model to .nn format for Unity embedding.
"""
tf.train.write_graph(self.graph,
self.cp_dir,
'raw_graph_def.pb',
as_text=False)
with self.graph.as_default():
target_nodes = ','.join(self._process_graph())
freeze_graph.freeze_graph(
input_graph=os.path.join(self.cp_dir, 'raw_graph_def.pb'),
input_binary=True,
input_checkpoint=tf.train.latest_checkpoint(self.cp_dir),
output_node_names=target_nodes,
output_graph=(os.path.join(self.cp_dir,
'frozen_graph_def.pb')),
# output_graph=(os.path.join(self.cp_dir,'model.bytes')),
clear_devices=True,
initializer_nodes='',
input_saver='',
restore_op_name='save/restore_all',
filename_tensor_name='save/Const:0')
tf2bc.convert(os.path.join(self.cp_dir, 'frozen_graph_def.pb'),
os.path.join(self.cp_dir, 'model.nn'))
def check_or_create(self, dicpath, name=''):
"""
check dictionary whether existing, if not then create it.
"""
if not os.path.exists(dicpath):
os.makedirs(dicpath)
print(f'create {name} directionary :', dicpath)
def close(self):
"""
end training, and export the training model
"""
self.export_model()
def get_global_step(self):
"""
get the current trianing step.
"""
return self.sess.run(self.global_step)
def set_global_step(self, num):
"""
set the start training step.
"""
with self.graph.as_default():
self.global_step.load(num, self.sess)
def plot_data(epi_num, logdir):
recorder = Recorder()
recorder.load(logdir)
recorder.plot_mpc_rl(epi_num)
eps=adam_epsilon)
scheduler = get_linear_schedule_with_warmup(
optimizer,
num_warmup_steps=warmup_steps,
num_training_steps=len(train_dataloader) * n_epoch)
# -------
# Initialize Recorder
# -------
SummaryPath = './Results/ALBERT-GLUE-%s/%s/%s/runs-%s-%d' % (
task_name.upper(), 'Prune' if args.prune else 'Quant', args.model_type,
'CR' if args.prune else 'bitW',
int(100.0 * args.CR) if args.prune else args.bitW)
if args.exp_spec is not None:
SummaryPath += ('-' + args.exp_spec)
recorder = Recorder(SummaryPath)
if recorder is not None:
recorder.write_arguments([args])
if args.first_eval:
result = evaluate(task_name, model, eval_dataloader,
model_type) # ['acc', 'f1', 'acc_f1']
print(result)
# --------------
# Begin Training
# --------------
for epoch_idx in range(n_epoch):
print('Epoch: %d' % epoch_idx)
model.train()
class HierarchicalMpc(object):
def __init__(self, task):
self.task = task
if self.task == 'left':
self.policy = LoadPolicy('G:\\env_build\\utils\\models\\left', 100000)
elif self.task == 'right':
self.policy = LoadPolicy('G:\\env_build\\utils\\models\\right', 145000)
elif self.task == 'straight':
self.policy = LoadPolicy('G:\\env_build\\utils\\models\\straight', 95000)
self.horizon = 25
self.num_future_data = 0
self.env = CrossroadEnd2end(training_task=self.task, num_future_data=self.num_future_data)
self.model = EnvironmentModel(self.task)
self.obs = self.env.reset()
self.stg = StaticTrajectoryGenerator_origin(mode='static_traj')
self.data2plot = []
self.mpc_cal_timer = TimerStat()
self.adp_cal_timer = TimerStat()
self.recorder = Recorder()
def reset(self):
self.obs = self.env.reset()
self.stg = StaticTrajectoryGenerator_origin(mode='static_traj')
self.recorder.reset()
self.recorder.save('.')
self.data2plot = []
return self.obs
def convert_vehs_to_abso(self, obs_rela):
ego_infos, tracking_infos, veh_rela = obs_rela[:6], \
obs_rela[6:6 + 3 * (1 + self.num_future_data)],\
obs_rela[6 + 3 * (1 + self.num_future_data):]
ego_vx, ego_vy, ego_r, ego_x, ego_y, ego_phi = ego_infos
ego = np.array([ego_x, ego_y, 0, 0] * int(len(veh_rela) / 4), dtype=np.float32)
vehs_abso = veh_rela + ego
out = np.concatenate((ego_infos, tracking_infos, vehs_abso), axis=0)
return out
def step(self):
traj_list, _ = self.stg.generate_traj(self.task, self.obs)
ADP_traj_return_value, MPC_traj_return_value = [], []
action_total = []
state_total = []
with self.mpc_cal_timer:
for ref_index, trajectory in enumerate(traj_list):
mpc = ModelPredictiveControl(self.horizon, self.task, self.num_future_data, ref_index)
state_all = np.array((list(self.obs[:6 + 3 * (1 + self.num_future_data)]) + [0, 0]) * self.horizon +
list(self.obs[:6 + 3 * (1 + self.num_future_data)])).reshape((-1, 1))
state, control, state_all, g_all, cost = mpc.mpc_solver(list(self.convert_vehs_to_abso(self.obs)), state_all)
state_total.append(state)
if any(g_all < -1):
print('optimization fail')
mpc_action = np.array([0., -1.])
state_all = np.array((list(self.obs[:9]) + [0, 0]) * self.horizon + list(self.obs[:9])).reshape(
(-1, 1))
else:
state_all = np.array((list(self.obs[:9]) + [0, 0]) * self.horizon + list(self.obs[:9])).reshape(
(-1, 1))
mpc_action = control[0]
MPC_traj_return_value.append(-cost.squeeze().tolist())
action_total.append(mpc_action)
MPC_traj_return_value = np.array(MPC_traj_return_value, dtype=np.float32)
MPC_path_index = np.argmax(MPC_traj_return_value)
MPC_action = action_total[MPC_path_index]
with self.adp_cal_timer:
for ref_index, trajectory in enumerate(traj_list):
self.env.set_traj(trajectory)
obs = self.env._get_obs()[np.newaxis, :]
traj_value = self.policy.values(obs)
ADP_traj_return_value.append(traj_value.numpy().squeeze().tolist())
ADP_traj_return_value = np.array(ADP_traj_return_value, dtype=np.float32)[:, 0]
ADP_path_index = np.argmax(ADP_traj_return_value)
if np.amax(ADP_traj_return_value) == np.amin(ADP_traj_return_value):
ADP_path_index = MPC_path_index
self.env.set_traj(traj_list[ADP_path_index])
self.obs_real = self.env._get_obs()
ADP_action = self.policy.run(self.obs_real).numpy()
self.recorder.record_compare(self.obs, ADP_action, MPC_action, self.adp_cal_timer.mean * 1000, self.mpc_cal_timer.mean * 1000,
ADP_path_index, MPC_path_index, 'both')
self.data2plot.append(dict(obs=self.obs,
ADP_action=ADP_action,
MPC_action=MPC_action,
ADP_path_index=ADP_path_index,
MPC_path_index=MPC_path_index,
mpc_time=self.mpc_cal_timer.mean * 1000,
))
self.obs, rew, done, _ = self.env.step(ADP_action)
self.render(traj_list, ADP_traj_return_value, ADP_path_index, MPC_traj_return_value, MPC_path_index, method='ADP')
state = state_total[MPC_path_index]
plt.plot([state[i][3] for i in range(1, self.horizon - 1)], [state[i][4] for i in range(1, self.horizon - 1)], 'r*')
plt.pause(0.001)
return done
def render(self, traj_list, ADP_traj_return_value, ADP_path_index, MPC_traj_return_value, MPC_path_index, method='ADP'):
square_length = CROSSROAD_SIZE
extension = 40
lane_width = LANE_WIDTH
light_line_width = 3
dotted_line_style = '--'
solid_line_style = '-'
plt.cla()
plt.title("Crossroad")
ax = plt.axes(xlim=(-square_length / 2 - extension, square_length / 2 + extension),
ylim=(-square_length / 2 - extension, square_length / 2 + extension))
plt.axis("equal")
plt.axis('off')
# ax.add_patch(plt.Rectangle((-square_length / 2, -square_length / 2),
# square_length, square_length, edgecolor='black', facecolor='none'))
ax.add_patch(plt.Rectangle((-square_length / 2 - extension, -square_length / 2 - extension),
square_length + 2 * extension, square_length + 2 * extension, edgecolor='black',
facecolor='none'))
# ----------arrow--------------
plt.arrow(lane_width / 2, -square_length / 2 - 10, 0, 5, color='b')
plt.arrow(lane_width / 2, -square_length / 2 - 10 + 5, -0.5, 0, color='b', head_width=1)
plt.arrow(lane_width * 1.5, -square_length / 2 - 10, 0, 5, color='b', head_width=1)
plt.arrow(lane_width * 2.5, -square_length / 2 - 10, 0, 5, color='b')
plt.arrow(lane_width * 2.5, -square_length / 2 - 10 + 5, 0.5, 0, color='b', head_width=1)
# ----------horizon--------------
plt.plot([-square_length / 2 - extension, -square_length / 2], [0, 0], color='black')
plt.plot([square_length / 2 + extension, square_length / 2], [0, 0], color='black')
#
for i in range(1, LANE_NUMBER + 1):
linestyle = dotted_line_style if i < LANE_NUMBER else solid_line_style
plt.plot([-square_length / 2 - extension, -square_length / 2], [i * lane_width, i * lane_width],
linestyle=linestyle, color='black')
plt.plot([square_length / 2 + extension, square_length / 2], [i * lane_width, i * lane_width],
linestyle=linestyle, color='black')
plt.plot([-square_length / 2 - extension, -square_length / 2], [-i * lane_width, -i * lane_width],
linestyle=linestyle, color='black')
plt.plot([square_length / 2 + extension, square_length / 2], [-i * lane_width, -i * lane_width],
linestyle=linestyle, color='black')
# ----------vertical----------------
plt.plot([0, 0], [-square_length / 2 - extension, -square_length / 2], color='black')
plt.plot([0, 0], [square_length / 2 + extension, square_length / 2], color='black')
#
for i in range(1, LANE_NUMBER + 1):
linestyle = dotted_line_style if i < LANE_NUMBER else solid_line_style
plt.plot([i * lane_width, i * lane_width], [-square_length / 2 - extension, -square_length / 2],
linestyle=linestyle, color='black')
plt.plot([i * lane_width, i * lane_width], [square_length / 2 + extension, square_length / 2],
linestyle=linestyle, color='black')
plt.plot([-i * lane_width, -i * lane_width], [-square_length / 2 - extension, -square_length / 2],
linestyle=linestyle, color='black')
plt.plot([-i * lane_width, -i * lane_width], [square_length / 2 + extension, square_length / 2],
linestyle=linestyle, color='black')
v_light = self.env.v_light
if v_light == 0:
v_color, h_color = 'green', 'red'
elif v_light == 1:
v_color, h_color = 'orange', 'red'
elif v_light == 2:
v_color, h_color = 'red', 'green'
else:
v_color, h_color = 'red', 'orange'
plt.plot([0, (LANE_NUMBER - 1) * lane_width], [-square_length / 2, -square_length / 2],
color=v_color, linewidth=light_line_width)
plt.plot([(LANE_NUMBER - 1) * lane_width, LANE_NUMBER * lane_width], [-square_length / 2, -square_length / 2],
color='green', linewidth=light_line_width)
plt.plot([-LANE_NUMBER * lane_width, -(LANE_NUMBER - 1) * lane_width], [square_length / 2, square_length / 2],
color='green', linewidth=light_line_width)
plt.plot([-(LANE_NUMBER - 1) * lane_width, 0], [square_length / 2, square_length / 2],
color=v_color, linewidth=light_line_width)
plt.plot([-square_length / 2, -square_length / 2], [0, -(LANE_NUMBER - 1) * lane_width],
color=h_color, linewidth=light_line_width)
plt.plot([-square_length / 2, -square_length / 2], [-(LANE_NUMBER - 1) * lane_width, -LANE_NUMBER * lane_width],
color='green', linewidth=light_line_width)
plt.plot([square_length / 2, square_length / 2], [(LANE_NUMBER - 1) * lane_width, 0],
color=h_color, linewidth=light_line_width)
plt.plot([square_length / 2, square_length / 2], [LANE_NUMBER * lane_width, (LANE_NUMBER - 1) * lane_width],
color='green', linewidth=light_line_width)
# ----------Oblique--------------
plt.plot([LANE_NUMBER * lane_width, square_length / 2], [-square_length / 2, -LANE_NUMBER * lane_width],
color='black')
plt.plot([LANE_NUMBER * lane_width, square_length / 2], [square_length / 2, LANE_NUMBER * lane_width],
color='black')
plt.plot([-LANE_NUMBER * lane_width, -square_length / 2], [-square_length / 2, -LANE_NUMBER * lane_width],
color='black')
plt.plot([-LANE_NUMBER * lane_width, -square_length / 2], [square_length / 2, LANE_NUMBER * lane_width],
color='black')
def is_in_plot_area(x, y, tolerance=5):
if -square_length / 2 - extension + tolerance < x < square_length / 2 + extension - tolerance and \
-square_length / 2 - extension + tolerance < y < square_length / 2 + extension - tolerance:
return True
else:
return False
def draw_rotate_rec(x, y, a, l, w, color, linestyle='-'):
RU_x, RU_y, _ = rotate_coordination(l / 2, w / 2, 0, -a)
RD_x, RD_y, _ = rotate_coordination(l / 2, -w / 2, 0, -a)
LU_x, LU_y, _ = rotate_coordination(-l / 2, w / 2, 0, -a)
LD_x, LD_y, _ = rotate_coordination(-l / 2, -w / 2, 0, -a)
ax.plot([RU_x + x, RD_x + x], [RU_y + y, RD_y + y], color=color, linestyle=linestyle)
ax.plot([RU_x + x, LU_x + x], [RU_y + y, LU_y + y], color=color, linestyle=linestyle)
ax.plot([LD_x + x, RD_x + x], [LD_y + y, RD_y + y], color=color, linestyle=linestyle)
ax.plot([LD_x + x, LU_x + x], [LD_y + y, LU_y + y], color=color, linestyle=linestyle)
def plot_phi_line(x, y, phi, color):
line_length = 5
x_forw, y_forw = x + line_length * cos(phi * pi / 180.), \
y + line_length * sin(phi * pi / 180.)
plt.plot([x, x_forw], [y, y_forw], color=color, linewidth=0.5)
# plot cars
for veh in self.env.all_vehicles:
veh_x = veh['x']
veh_y = veh['y']
veh_phi = veh['phi']
veh_l = veh['l']
veh_w = veh['w']
if is_in_plot_area(veh_x, veh_y):
plot_phi_line(veh_x, veh_y, veh_phi, 'black')
draw_rotate_rec(veh_x, veh_y, veh_phi, veh_l, veh_w, 'black')
# plot_interested vehs
# for mode, num in self.veh_mode_dict.items():
# for i in range(num):
# veh = self.interested_vehs[mode][i]
# veh_x = veh['x']
# veh_y = veh['y']
# veh_phi = veh['phi']
# veh_l = veh['l']
# veh_w = veh['w']
# task2color = {'left': 'b', 'straight': 'c', 'right': 'm'}
#
# if is_in_plot_area(veh_x, veh_y):
# plot_phi_line(veh_x, veh_y, veh_phi, 'black')
# task = MODE2TASK[mode]
# color = task2color[task]
# draw_rotate_rec(veh_x, veh_y, veh_phi, veh_l, veh_w, color, linestyle=':')
ego_v_x = self.env.ego_dynamics['v_x']
ego_v_y = self.env.ego_dynamics['v_y']
ego_r = self.env.ego_dynamics['r']
ego_x = self.env.ego_dynamics['x']
ego_y = self.env.ego_dynamics['y']
ego_phi = self.env.ego_dynamics['phi']
ego_l = self.env.ego_dynamics['l']
ego_w = self.env.ego_dynamics['w']
ego_alpha_f = self.env.ego_dynamics['alpha_f']
ego_alpha_r = self.env.ego_dynamics['alpha_r']
alpha_f_bound = self.env.ego_dynamics['alpha_f_bound']
alpha_r_bound = self.env.ego_dynamics['alpha_r_bound']
r_bound = self.env.ego_dynamics['r_bound']
plot_phi_line(ego_x, ego_y, ego_phi, 'fuchsia')
draw_rotate_rec(ego_x, ego_y, ego_phi, ego_l, ego_w, 'fuchsia')
# plot future data
tracking_info = self.obs[
self.env.ego_info_dim:self.env.ego_info_dim + self.env.per_tracking_info_dim * (self.env.num_future_data + 1)]
future_path = tracking_info[self.env.per_tracking_info_dim:]
for i in range(self.env.num_future_data):
delta_x, delta_y, delta_phi = future_path[i * self.env.per_tracking_info_dim:
(i + 1) * self.env.per_tracking_info_dim]
path_x, path_y, path_phi = ego_x + delta_x, ego_y + delta_y, ego_phi - delta_phi
plt.plot(path_x, path_y, 'g.')
plot_phi_line(path_x, path_y, path_phi, 'g')
delta_, _, _ = tracking_info[:3]
indexs, points = self.env.ref_path.find_closest_point(np.array([ego_x], np.float32), np.array([ego_y], np.float32))
path_x, path_y, path_phi = points[0][0], points[1][0], points[2][0]
# plt.plot(path_x, path_y, 'g.')
delta_x, delta_y, delta_phi = ego_x - path_x, ego_y - path_y, ego_phi - path_phi
# plot real time traj
try:
color = ['blue', 'coral', 'cyan']
for i, item in enumerate(traj_list):
if method == 'ADP':
if i == ADP_path_index:
plt.plot(item.path[0], item.path[1], color=color[i])
else:
plt.plot(item.path[0], item.path[1], color=color[i], alpha=0.3)
indexs, points = item.find_closest_point(np.array([ego_x], np.float32), np.array([ego_y], np.float32))
path_x, path_y, path_phi = points[0][0], points[1][0], points[2][0]
plt.plot(path_x, path_y, color=color[i])
elif method == 'MPC':
if i == MPC_path_index:
plt.plot(item.path[0], item.path[1], color=color[i])
else:
plt.plot(item.path[0], item.path[1], color=color[i], alpha=0.3)
indexs, points = item.find_closest_point(np.array([ego_x], np.float32), np.array([ego_y], np.float32))
path_x, path_y, path_phi = points[0][0], points[1][0], points[2][0]
plt.plot(path_x, path_y, color=color[i])
except Exception:
pass
# plot ego dynamics
text_x, text_y_start = -120, 60
ge = iter(range(0, 1000, 4))
plt.text(text_x, text_y_start - next(ge), 'ego_x: {:.2f}m'.format(ego_x))
plt.text(text_x, text_y_start - next(ge), 'ego_y: {:.2f}m'.format(ego_y))
plt.text(text_x, text_y_start - next(ge), 'path_x: {:.2f}m'.format(path_x))
plt.text(text_x, text_y_start - next(ge), 'path_y: {:.2f}m'.format(path_y))
plt.text(text_x, text_y_start - next(ge), 'delta_: {:.2f}m'.format(delta_))
plt.text(text_x, text_y_start - next(ge), 'delta_x: {:.2f}m'.format(delta_x))
plt.text(text_x, text_y_start - next(ge), 'delta_y: {:.2f}m'.format(delta_y))
plt.text(text_x, text_y_start - next(ge), r'ego_phi: ${:.2f}\degree$'.format(ego_phi))
plt.text(text_x, text_y_start - next(ge), r'path_phi: ${:.2f}\degree$'.format(path_phi))
plt.text(text_x, text_y_start - next(ge), r'delta_phi: ${:.2f}\degree$'.format(delta_phi))
plt.text(text_x, text_y_start - next(ge), 'v_x: {:.2f}m/s'.format(ego_v_x))
plt.text(text_x, text_y_start - next(ge), 'exp_v: {:.2f}m/s'.format(self.env.exp_v))
plt.text(text_x, text_y_start - next(ge), 'v_y: {:.2f}m/s'.format(ego_v_y))
plt.text(text_x, text_y_start - next(ge), 'yaw_rate: {:.2f}rad/s'.format(ego_r))
plt.text(text_x, text_y_start - next(ge), 'yaw_rate bound: [{:.2f}, {:.2f}]'.format(-r_bound, r_bound))
plt.text(text_x, text_y_start - next(ge), r'$\alpha_f$: {:.2f} rad'.format(ego_alpha_f))
plt.text(text_x, text_y_start - next(ge), r'$\alpha_f$ bound: [{:.2f}, {:.2f}] '.format(-alpha_f_bound,
alpha_f_bound))
plt.text(text_x, text_y_start - next(ge), r'$\alpha_r$: {:.2f} rad'.format(ego_alpha_r))
plt.text(text_x, text_y_start - next(ge), r'$\alpha_r$ bound: [{:.2f}, {:.2f}] '.format(-alpha_r_bound,
alpha_r_bound))
if self.env.action is not None:
steer, a_x = self.env.action[0], self.env.action[1]
plt.text(text_x, text_y_start - next(ge),
r'steer: {:.2f}rad (${:.2f}\degree$)'.format(steer, steer * 180 / np.pi))
plt.text(text_x, text_y_start - next(ge), 'a_x: {:.2f}m/s^2'.format(a_x))
text_x, text_y_start = 70, 60
ge = iter(range(0, 1000, 4))
# done info
plt.text(text_x, text_y_start - next(ge), 'done info: {}'.format(self.env.done_type))
# reward info
if self.env.reward_info is not None:
for key, val in self.env.reward_info.items():
plt.text(text_x, text_y_start - next(ge), '{}: {:.4f}'.format(key, val))
# indicator for Atrajectory selection
text_x, text_y_start = 18, -70
ge = iter(range(0, 1000, 6))
plt.text(text_x+10, text_y_start - next(ge), 'ADP', fontsize=14, color='r', fontstyle='italic')
if ADP_traj_return_value is not None:
for i, value in enumerate(ADP_traj_return_value):
if i == ADP_path_index:
plt.text(text_x, text_y_start - next(ge), 'Path cost={:.4f}'.format(value), fontsize=14,
color=color[i], fontstyle='italic')
else:
plt.text(text_x, text_y_start - next(ge), 'Path cost={:.4f}'.format(value), fontsize=12,
color=color[i], fontstyle='italic')
text_x, text_y_start = -36, -70
ge = iter(range(0, 1000, 6))
plt.text(text_x+10, text_y_start - next(ge), 'MPC', fontsize=14, color='r', fontstyle='italic')
if MPC_traj_return_value is not None:
for i, value in enumerate(MPC_traj_return_value):
if i == MPC_path_index:
plt.text(text_x, text_y_start - next(ge), 'Path cost={:.4f}'.format(value), fontsize=14,
color=color[i], fontstyle='italic')
else:
plt.text(text_x, text_y_start - next(ge), 'Path cost={:.4f}'.format(value), fontsize=12,
color=color[i], fontstyle='italic')
plt.pause(0.001)
target_summary_path = '%s/runs-%s/CR%.2f' %(save_root, target_dataset_name, 100 * overall_CR)
for SummaryPath in [source_summary_path, target_summary_path]:
if args.exp_spec is not '':
SummaryPath += ('-' + args.exp_spec)
if os.path.exists(SummaryPath):
print('Record exist, remove')
input()
shutil.rmtree(SummaryPath)
os.makedirs(SummaryPath)
else:
os.makedirs(SummaryPath)
source_recorder = Recorder(SummaryPath=source_summary_path, dataset_name=source_dataset_name)
target_recorder = Recorder(SummaryPath=target_summary_path, dataset_name=source_dataset_name)
alpha_change_point_file = open('%s/alpha_change_point.txt' %(target_summary_path), 'w+')
##################
# Begin Training #
##################
best_acc_list = [] # Best test acc in each training period under various alpha
niter = 0
for ite, CR_ratio in enumerate(CR_list):
alpha = alpha_list[ite]
print('Adaptive iteration: %d, alpha: %.3f' % (ite, alpha))
print('Current CR: %s' % CR_ratio)
print('niter: %d' %niter)
class WebBrowser(QObject):
dic_url = {'SWGS': 'http://172.31.196.21:8060/websquare/websquare.html?w2xPath=/SWING/lib/xml/ZLIBSMAN90010.xml&coClCd=T&svrLocation=SWGS'}
browser_log = []
requestList = []
eventList = []
browser_log_hst = []
HOME = os.path.expanduser("~")
SAVE_PATH = os.path.join(HOME, 'Test_Tool')
HOME_SAVE_PATH = os.path.join(HOME, '.test_tool')
def __init__(self, sid, url='', websocket_server=None):
QObject.__init__(self)
self.sid = sid
self.url = url
self.driver = None
self.recorder = None
self.requestList = []
self.eventList = []
self.browser_log = []
self.browser_log_hst = []
self.chromedriver_version = ''
self.recording = False
self.web_iframe_list = []
# Load setting in the main thread
self.settings = Settings(self.HOME_SAVE_PATH)
self.settings.load()
self.settings.save()
self._loadSetting()
self.config = Config()
self.paths = [r"C:\Program Files\Google\Chrome\Application\chrome.exe",
r"C:\Program Files (x86)\Google\Chrome\Application\chrome.exe"]
self.paths.extend(self.config.getlist("section_browser", "BROWSER_PATH"))
self.browser_version = self.config.get('section_browser', 'BROWSER_VER')
if self.sid in self.dic_url.keys():
print(self.dic_url[self.sid])
elif self.sid == 'URL':
print(self.url)
self.dic_url[self.sid] = self.url
else:
print('미존재 SID')
self.websocket_server = websocket_server
self.websocket_server.receivedSignal.connect(self.websocketSeverReceived)
def _loadSetting(self):
'''
설정값으로 재설정
:return: None
'''
self.getDriverVersion()
def getDriverVersion(self):
self.settings.load()
self.chromedriver_version = self.settings.get("CHROMEDRIVER_VERSION", "")
def getVersionViaCom(self):
version_info = []
paths = self.paths
try:
parser = Dispatch("Scripting.FileSystemObject")
except:
if self.browser_version:
result_verion= self.browser_version
else:
result_verion = ''
return result_verion
for file_name in paths:
try:
version = parser.GetFileVersion(file_name)
version_info.append(version)
except:
pass
try:
result_verion = version_info[0].split('.')[0]
except:
result_verion = ''
return result_verion
def processBrowserLogEntry(self, entry):
response = json.loads(entry['message'])['message']
return response
def findEventLog(self, type, evets):
requestList = []
xhr_events = [event for event in evets if 'type' in event['params'].keys() and event['params']['type'] == type]
xhr_events = [event for event in xhr_events if event['params']['request']['method'] == 'POST' and 'postData' in event['params']['request'].keys()]
#print('==== XHR List ====')
for idx, xhr in enumerate(xhr_events):
post_data = xhr['params']['request']['postData']
try:
trx_code = json.loads(post_data)['HEAD']['Trx_Code']
except json.JSONDecodeError as e:
print(e)
print(post_data)
request_id = xhr['params']['requestId']
#print('{index} - {trxCode} - {requestId}'.format(index=idx, trxCode=trx_code, requestId=request_id))
inputData = {}
outputData = {}
# request결과가 없는 경우가 있음 (ZNGMSATH10060_TR03) 결과가 없으면 제외
try:
inputData = json.loads(self.driver.execute_cdp_cmd('Network.getRequestPostData', {'requestId': str(request_id)})['postData'])
outputData = json.loads(self.driver.execute_cdp_cmd('Network.getResponseBody', {'requestId': str(request_id)})['body'])
if 'dataInfo' not in list(outputData.keys()):
outputData = makeDataInfo(outputData)
requestDtl = {'trx_code': trx_code,
'request_id': request_id,
'input_data': inputData,
'output_data': outputData}
requestList.append(requestDtl)
except:
pass
return requestList
def getInputDataset(self, events, idx):
return json.loads(events[idx]['params']['request']['postData'])
def popUp(self):
version = self.getVersionViaCom()
chromedriver_path = os.path.join(self.SAVE_PATH, 'chromedriver.exe')
self.getDriverVersion()
if not os.path.exists(chromedriver_path):
print('chromedriver 파일 미존재 - 생성....')
self.downloadChromedriver(version, chromedriver_path)
elif version == self.chromedriver_version:
print('동일한 Chrome / chromedriver 버전 - [{}]'.format(version))
else:
print('Chrome [{chrome_version}] / chromedriver [{chromedriver_version}] 버전이 다른 경우'.format(chrome_version=version, chromedriver_version=self.chromedriver_version))
self.downloadChromedriver(version, chromedriver_path)
self.web_iframe_list = []
chrome_options = Options()
#chrome_options.add_argument("start-maximized")
chrome_options.add_argument("disable-infobars")
chrome_options.add_argument("--disable-extensions")
chrome_options.add_argument("--no-sandbox")
chrome_options.add_argument("--disable-gpu")
chrome_options.add_argument("window-size=1920x1080") # 가져올 크기를 결정
chrome_options.set_capability('unhandledPromptBehavior', 'accept')
caps = DesiredCapabilities.CHROME
caps['goog:loggingPrefs'] = {'performance': 'ALL'}
if self.driver and self.getStatus():
pass
else:
# 드라이버 객체 생성
self.driver = webdriver.Chrome(chromedriver_path, chrome_options=chrome_options, desired_capabilities=caps)
#self.driver.implicitly_wait(3) # 드라이버 초기화를 위해 3초 대기
self.get()
def get(self):
self.driver.get(self.dic_url[self.sid])
if self.sid == 'SWGS':
try:
self.driver.switch_to_alert()
alert = WebDriverWait(self.driver, 1).until(EC.alert_is_present())
alert.accept()
except NoAlertPresentException:
pass
finally:
while True:
try:
WebDriverWait(self.driver, 0).until(EC.element_to_be_clickable((By.ID, 'bodyBlock')))
except:
break
try:
WebDriverWait(self.driver, 60).until(EC.presence_of_element_located((By.ID, 'mdi01_subWindow0_iframe')))
finally:
print('Page is ready!')
self.driver.execute_script('''top.document.title = "(FOR AUTO TEST TOOL)"''')
# Recorder 선언
self.recorder = Recorder(self.driver)
self.recorder.addEventListener()
#threading.Timer(3, self.getBrowserEvents).start()
def getRequest(self):
try:
self.driver.switch_to_alert()
except NoAlertPresentException:
log = self.driver.get_log('performance')
self.browser_log.extend(log)
self.browser_log_hst.extend(log)
events = [self.processBrowserLogEntry(entry) for entry in self.browser_log]
# response_events = [event for event in events if 'Network.response' in event['method']]
request_events = [event for event in events if 'Network.requestWillBeSent' in event['method']]
self.requestList = self.findEventLog('XHR', request_events)
return self.requestList
def getRequestHst(self):
try:
self.driver.switch_to_alert()
except NoAlertPresentException:
log = self.driver.get_log('performance')
self.browser_log.extend(log)
self.browser_log_hst.extend(log)
events = [self.processBrowserLogEntry(entry) for entry in self.browser_log_hst]
# response_events = [event for event in events if 'Network.response' in event['method']]
request_events = [event for event in events if 'Network.requestWillBeSent' in event['method']]
self.requestHst = self.findEventLog('XHR', request_events)
return self.requestHst
def clear(self):
self.browser_log_hst.extend(self.driver.get_log('performance'))
self.browser_log = []
self.requestList = []
print('Request Clear Successful')
def clearUiEventList(self):
self.recorder.clearEventList()
def recordUiEvent(self):
if not self.recording:
self.recorder.clearEventList()
self.recorder.addEventListener()
self.recording = not self.recording
def websocketSeverReceived(self, data):
'''
:websocket Server로부터 data를 받았을때 발생하는 이벤트
data: (str) 'add_iframe'
'''
#print(data)
if data == 'add_iframe':
self.recorder.addEventListener()
def getBrowserEvents(self):
events = []
try:
events = self.recorder.getEventList()
#threading.Timer(3, self.getBrowserEvents).start()
except JavascriptException:
self.recorder.addEventListener()
events = self.recorder.getEventList()
#threading.Timer(3, self.getBrowserEvents).start()
except WebDriverException:
print('Browser 종료')
finally:
#print(events)
self.recorder.addEventListener()
return events
# if events:
# print(events)
# self.eventList.extend(events)
#self.recorder.clearEventList()
# for event in events:
# # One-time initialization
# toaster = ToastNotifier()
#
# # Show notification whenever needed
# toaster.show_toast(title="Command Record",
# msg="Command:{command}\nTarget:{target}\nValue:{value}".format(command=event[1], target=event[3], value=event[5]),
# threaded=True,
# icon_path='auto_test_main.ico',
# duration=1)
# notification.notify(
# title='Command Record',
# ticker='A',
# message="Command:{command}\nTarget:{target}\nValue:{value}".format(command=event[1], target=event[3], value=event[5]),
# app_name="Auto Test Tool 0.0.4",
# app_icon='auto_test_main.ico',
# timeout=5, # seconds
# )
def getDriver(self):
return self.driver
def getEventList(self):
return self.eventList
def getSessionId(self):
if self.driver:
return self.driver.session_id
else:
return False
def setSessionId(self, session_id):
self.driver.session_id = session_id
def getStatus(self):
try:
self.driver.window_handles
return True
except:
return False
def setFocus(self):
self.driver.switch_to_window(self.driver.current_window_handle)
def saveChromedriverVersion(self, version):
'''
chromedriver version 정보를 저장
'''
self.settings.load()
self.settings["CHROMEDRIVER_VERSION"] = version
self.settings.save()
def chromedriverWebOn(self):
'''
chromedriver site 접속 가능여부 체크
'''
url = "https://chromedriver.chromium.org/downloads"
try:
res = urllib.request.urlopen(url)
if res.status == 200:
return True
else:
return False
except:
return False
def downloadChromedriver(self, version, chromedriver_path):
'''
chromedriver가 미존재하거나 버전이 다른 경우 발생하는 이벤트
- 온라인 연결이 가능한 경우는 다운로드
- 온라인 연결이 불가능한 경우 임시 파일 중 가능한 버전으로 대체
:param version: (str) '81'
:param chromedriver_path: (str)
'''
if self.chromedriverWebOn():
print("Chromedriver Downloading...")
download_url = self.findDriverVersion(version)
download_path = os.path.join(self.SAVE_PATH, "chromedriver.zip")
urllib.request.urlretrieve(download_url, download_path)
print("Download Complete!")
proc_name = "chromedriver.exe"
# 현재 수행중인 chromedriver 존재여부 확인
for proc in psutil.process_iter():
try:
if proc.name().lower() == proc_name.lower():
print('현재 수행중인 chromedriver kill [{}]'.format(proc.name()))
proc.kill()
except (psutil.NoSuchProcess, psutil.AccessDenied, psutil.ZombieProcess):
pass
try:
with zipfile.ZipFile(download_path) as zf:
zf.extractall(path=self.SAVE_PATH)
os.remove(download_path)
except Exception as e:
print(e)
self.saveChromedriverVersion(version)
else:
print('Online 연결 불가능 - 오프라인 파일 생성')
try:
if version == '81':
shutil.copy('chromedriver_tmp/chromedriver_81.exe', chromedriver_path)
self.saveChromedriverVersion(version)
elif version == '80':
shutil.copy('chromedriver_tmp/chromedriver_80.exe', chromedriver_path)
self.saveChromedriverVersion(version)
elif version == '79':
shutil.copy('chromedriver_tmp/chromedriver_79.exe', chromedriver_path)
self.saveChromedriverVersion(version)
elif version in ['65', '66', '67']:
shutil.copy('chromedriver_tmp/chromedriver_65.exe', chromedriver_path)
self.saveChromedriverVersion(version)
else:
print('Chrome Version Check : {} - 오프라인은 65, 66, 67, 79, 80, 81 Version만 지원가능'.format(version))
return False
except:
pass
pass
def findDriverVersion(self, find_version):
'''
chromedriver를 다운로드 할 url를 찾아 string으로 Return
chromedriver chrome
85 85
. .
. .
71 71
70 70
---------------------
2.46 71-73
2.45 70-72
2.44 69-71
2.43 69-71
2.42 68-70
2.41 67-69
2.40 66-68
2.39 66-68
2.38 65-67
2.37 64-66
2.36 63-65
2.35 62-64
2.34 61-63
2.33 60-62
---------------------
2.28 57+
2.25 54+
2.24 53+
2.22 51+
2.19 44+
2.15 42+
:param find_version: (str) '81'
:return: (str) https://chromedriver.storage.googleapis.com/81.0.4044.138/chromedriver_win32.zip
'''
download_url = ''
base_url = "https://chromedriver.chromium.org/downloads"
base_req = requests.get(base_url)
base_soup = BeautifulSoup(base_req.content, "html.parser")
atags = base_soup.select("a")
for a in atags:
m = re.compile("ChromeDriver (.*)")
p = m.search(a.text)
try:
version = p.group(1)
try:
m = re.compile("(\d*)\..*")
p = m.search(version)
if int(find_version) == int(p.group(1)):
download_url = "/".join(
[
"https://chromedriver.storage.googleapis.com",
version,
"chromedriver_win32.zip",
]
)
break
except:
pass
except AttributeError:
pass
return download_url
class HierarchicalDecision(object):
def __init__(self, task, train_exp_dir, ite, logdir=None):
self.task = task
self.policy = LoadPolicy('../utils/models/{}/{}'.format(task, train_exp_dir), ite)
self.env = CrossroadEnd2end(training_task=self.task, mode='testing')
self.model = EnvironmentModel(self.task, mode='selecting')
self.recorder = Recorder()
self.episode_counter = -1
self.step_counter = -1
self.obs = None
self.stg = MultiPathGenerator()
self.step_timer = TimerStat()
self.ss_timer = TimerStat()
self.logdir = logdir
if self.logdir is not None:
config = dict(task=task, train_exp_dir=train_exp_dir, ite=ite)
with open(self.logdir + '/config.json', 'w', encoding='utf-8') as f:
json.dump(config, f, ensure_ascii=False, indent=4)
self.fig = plt.figure(figsize=(8, 8))
plt.ion()
self.hist_posi = []
self.old_index = 0
self.path_list = self.stg.generate_path(self.task)
# ------------------build graph for tf.function in advance-----------------------
for i in range(3):
obs = self.env.reset()
obs = tf.convert_to_tensor(obs[np.newaxis, :], dtype=tf.float32)
self.is_safe(obs, i)
obs = self.env.reset()
obs_with_specific_shape = np.tile(obs, (3, 1))
self.policy.run_batch(obs_with_specific_shape)
self.policy.obj_value_batch(obs_with_specific_shape)
# ------------------build graph for tf.function in advance-----------------------
self.reset()
def reset(self,):
self.obs = self.env.reset()
self.recorder.reset()
self.old_index = 0
self.hist_posi = []
if self.logdir is not None:
self.episode_counter += 1
os.makedirs(self.logdir + '/episode{}/figs'.format(self.episode_counter))
self.step_counter = -1
self.recorder.save(self.logdir)
if self.episode_counter >= 1:
select_and_rename_snapshots_of_an_episode(self.logdir, self.episode_counter-1, 12)
self.recorder.plot_and_save_ith_episode_curves(self.episode_counter-1,
self.logdir + '/episode{}/figs'.format(self.episode_counter-1),
isshow=False)
return self.obs
# @tf.function
# def is_safe(self, obs, path_index):
# self.model.ref_path.set_path(path_index)
# action = self.policy.run_batch(obs)
# veh2veh4real = self.model.ss(obs, action, lam=0.1)
# return False if veh2veh4real[0] > 0 else True
@tf.function
def is_safe(self, obs, path_index):
self.model.add_traj(obs, path_index)
punish = 0.
for step in range(5):
action = self.policy.run_batch(obs)
obs, _, _, _, veh2veh4real, _ = self.model.rollout_out(action)
punish += veh2veh4real[0]
return False if punish > 0 else True
def safe_shield(self, real_obs, path_index):
action_safe_set = [[[0., -1.]]]
real_obs = tf.convert_to_tensor(real_obs[np.newaxis, :], dtype=tf.float32)
obs = real_obs
if not self.is_safe(obs, path_index):
print('SAFETY SHIELD STARTED!')
return np.array(action_safe_set[0], dtype=np.float32).squeeze(0), True
else:
return self.policy.run_batch(real_obs).numpy()[0], False
def step(self):
self.step_counter += 1
with self.step_timer:
obs_list = []
# select optimal path
for path in self.path_list:
self.env.set_traj(path)
obs_list.append(self.env._get_obs())
all_obs = tf.stack(obs_list, axis=0)
path_values = self.policy.obj_value_batch(all_obs).numpy()
old_value = path_values[self.old_index]
new_index, new_value = int(np.argmin(path_values)), min(path_values) # value is to approximate (- sum of reward)
path_index = self.old_index if old_value - new_value < 0.1 else new_index
self.old_index = path_index
self.env.set_traj(self.path_list[path_index])
self.obs_real = obs_list[path_index]
# obtain safe action
with self.ss_timer:
safe_action, is_ss = self.safe_shield(self.obs_real, path_index)
print('ALL TIME:', self.step_timer.mean, 'ss', self.ss_timer.mean)
self.render(self.path_list, path_values, path_index)
self.recorder.record(self.obs_real, safe_action, self.step_timer.mean,
path_index, path_values, self.ss_timer.mean, is_ss)
self.obs, r, done, info = self.env.step(safe_action)
return done
def render(self, traj_list, path_values, path_index):
square_length = CROSSROAD_SIZE
extension = 40
lane_width = LANE_WIDTH
light_line_width = 3
dotted_line_style = '--'
solid_line_style = '-'
plt.cla()
ax = plt.axes([-0.05, -0.05, 1.1, 1.1])
for ax in self.fig.get_axes():
ax.axis('off')
ax.axis("equal")
# ----------arrow--------------
plt.arrow(lane_width / 2, -square_length / 2 - 10, 0, 5, color='b')
plt.arrow(lane_width / 2, -square_length / 2 - 10 + 5, -0.5, 0, color='b', head_width=1)
plt.arrow(lane_width * 1.5, -square_length / 2 - 10, 0, 4, color='b', head_width=1)
plt.arrow(lane_width * 2.5, -square_length / 2 - 10, 0, 5, color='b')
plt.arrow(lane_width * 2.5, -square_length / 2 - 10 + 5, 0.5, 0, color='b', head_width=1)
# ----------horizon--------------
plt.plot([-square_length / 2 - extension, -square_length / 2], [0.3, 0.3], color='orange')
plt.plot([-square_length / 2 - extension, -square_length / 2], [-0.3, -0.3], color='orange')
plt.plot([square_length / 2 + extension, square_length / 2], [0.3, 0.3], color='orange')
plt.plot([square_length / 2 + extension, square_length / 2], [-0.3, -0.3], color='orange')
#
for i in range(1, LANE_NUMBER + 1):
linestyle = dotted_line_style if i < LANE_NUMBER else solid_line_style
linewidth = 1 if i < LANE_NUMBER else 2
plt.plot([-square_length / 2 - extension, -square_length / 2], [i * lane_width, i * lane_width],
linestyle=linestyle, color='black', linewidth=linewidth)
plt.plot([square_length / 2 + extension, square_length / 2], [i * lane_width, i * lane_width],
linestyle=linestyle, color='black', linewidth=linewidth)
plt.plot([-square_length / 2 - extension, -square_length / 2], [-i * lane_width, -i * lane_width],
linestyle=linestyle, color='black', linewidth=linewidth)
plt.plot([square_length / 2 + extension, square_length / 2], [-i * lane_width, -i * lane_width],
linestyle=linestyle, color='black', linewidth=linewidth)
# ----------vertical----------------
plt.plot([0.3, 0.3], [-square_length / 2 - extension, -square_length / 2], color='orange')
plt.plot([-0.3, -0.3], [-square_length / 2 - extension, -square_length / 2], color='orange')
plt.plot([0.3, 0.3], [square_length / 2 + extension, square_length / 2], color='orange')
plt.plot([-0.3, -0.3], [square_length / 2 + extension, square_length / 2], color='orange')
#
for i in range(1, LANE_NUMBER + 1):
linestyle = dotted_line_style if i < LANE_NUMBER else solid_line_style
linewidth = 1 if i < LANE_NUMBER else 2
plt.plot([i * lane_width, i * lane_width], [-square_length / 2 - extension, -square_length / 2],
linestyle=linestyle, color='black', linewidth=linewidth)
plt.plot([i * lane_width, i * lane_width], [square_length / 2 + extension, square_length / 2],
linestyle=linestyle, color='black', linewidth=linewidth)
plt.plot([-i * lane_width, -i * lane_width], [-square_length / 2 - extension, -square_length / 2],
linestyle=linestyle, color='black', linewidth=linewidth)
plt.plot([-i * lane_width, -i * lane_width], [square_length / 2 + extension, square_length / 2],
linestyle=linestyle, color='black', linewidth=linewidth)
v_light = self.env.v_light
if v_light == 0:
v_color, h_color = 'green', 'red'
elif v_light == 1:
v_color, h_color = 'orange', 'red'
elif v_light == 2:
v_color, h_color = 'red', 'green'
else:
v_color, h_color = 'red', 'orange'
plt.plot([0, (LANE_NUMBER - 1) * lane_width], [-square_length / 2, -square_length / 2],
color=v_color, linewidth=light_line_width)
plt.plot([(LANE_NUMBER - 1) * lane_width, LANE_NUMBER * lane_width], [-square_length / 2, -square_length / 2],
color='green', linewidth=light_line_width)
plt.plot([-LANE_NUMBER * lane_width, -(LANE_NUMBER - 1) * lane_width], [square_length / 2, square_length / 2],
color='green', linewidth=light_line_width)
plt.plot([-(LANE_NUMBER - 1) * lane_width, 0], [square_length / 2, square_length / 2],
color=v_color, linewidth=light_line_width)
plt.plot([-square_length / 2, -square_length / 2], [0, -(LANE_NUMBER - 1) * lane_width],
color=h_color, linewidth=light_line_width)
plt.plot([-square_length / 2, -square_length / 2], [-(LANE_NUMBER - 1) * lane_width, -LANE_NUMBER * lane_width],
color='green', linewidth=light_line_width)
plt.plot([square_length / 2, square_length / 2], [(LANE_NUMBER - 1) * lane_width, 0],
color=h_color, linewidth=light_line_width)
plt.plot([square_length / 2, square_length / 2], [LANE_NUMBER * lane_width, (LANE_NUMBER - 1) * lane_width],
color='green', linewidth=light_line_width)
# ----------Oblique--------------
plt.plot([LANE_NUMBER * lane_width, square_length / 2], [-square_length / 2, -LANE_NUMBER * lane_width],
color='black', linewidth=2)
plt.plot([LANE_NUMBER * lane_width, square_length / 2], [square_length / 2, LANE_NUMBER * lane_width],
color='black', linewidth=2)
plt.plot([-LANE_NUMBER * lane_width, -square_length / 2], [-square_length / 2, -LANE_NUMBER * lane_width],
color='black', linewidth=2)
plt.plot([-LANE_NUMBER * lane_width, -square_length / 2], [square_length / 2, LANE_NUMBER * lane_width],
color='black', linewidth=2)
def is_in_plot_area(x, y, tolerance=5):
if -square_length / 2 - extension + tolerance < x < square_length / 2 + extension - tolerance and \
-square_length / 2 - extension + tolerance < y < square_length / 2 + extension - tolerance:
return True
else:
return False
def draw_rotate_rec(x, y, a, l, w, c):
bottom_left_x, bottom_left_y, _ = rotate_coordination(-l / 2, w / 2, 0, -a)
ax.add_patch(plt.Rectangle((x + bottom_left_x, y + bottom_left_y), w, l, edgecolor=c,
facecolor='white', angle=-(90 - a), zorder=50))
def plot_phi_line(x, y, phi, color):
line_length = 3
x_forw, y_forw = x + line_length * cos(phi * pi / 180.), \
y + line_length * sin(phi * pi / 180.)
plt.plot([x, x_forw], [y, y_forw], color=color, linewidth=0.5)
# plot cars
for veh in self.env.all_vehicles:
veh_x = veh['x']
veh_y = veh['y']
veh_phi = veh['phi']
veh_l = veh['l']
veh_w = veh['w']
if is_in_plot_area(veh_x, veh_y):
plot_phi_line(veh_x, veh_y, veh_phi, 'black')
draw_rotate_rec(veh_x, veh_y, veh_phi, veh_l, veh_w, 'black')
# plot_interested vehs
# for mode, num in self.env.veh_mode_dict.items():
# for i in range(num):
# veh = self.env.interested_vehs[mode][i]
# veh_x = veh['x']
# veh_y = veh['y']
# veh_phi = veh['phi']
# veh_l = veh['l']
# veh_w = veh['w']
# task2color = {'left': 'b', 'straight': 'c', 'right': 'm'}
#
# if is_in_plot_area(veh_x, veh_y):
# plot_phi_line(veh_x, veh_y, veh_phi, 'black')
# task = MODE2TASK[mode]
# color = task2color[task]
# draw_rotate_rec(veh_x, veh_y, veh_phi, veh_l, veh_w, color)
ego_v_x = self.env.ego_dynamics['v_x']
ego_v_y = self.env.ego_dynamics['v_y']
ego_r = self.env.ego_dynamics['r']
ego_x = self.env.ego_dynamics['x']
ego_y = self.env.ego_dynamics['y']
ego_phi = self.env.ego_dynamics['phi']
ego_l = self.env.ego_dynamics['l']
ego_w = self.env.ego_dynamics['w']
ego_alpha_f = self.env.ego_dynamics['alpha_f']
ego_alpha_r = self.env.ego_dynamics['alpha_r']
alpha_f_bound = self.env.ego_dynamics['alpha_f_bound']
alpha_r_bound = self.env.ego_dynamics['alpha_r_bound']
r_bound = self.env.ego_dynamics['r_bound']
plot_phi_line(ego_x, ego_y, ego_phi, 'fuchsia')
draw_rotate_rec(ego_x, ego_y, ego_phi, ego_l, ego_w, 'fuchsia')
self.hist_posi.append((ego_x, ego_y))
# plot history
xs = [pos[0] for pos in self.hist_posi]
ys = [pos[1] for pos in self.hist_posi]
plt.scatter(np.array(xs), np.array(ys), color='fuchsia', alpha=0.1)
# plot future data
tracking_info = self.obs[
self.env.ego_info_dim:self.env.ego_info_dim + self.env.per_tracking_info_dim * (self.env.num_future_data + 1)]
future_path = tracking_info[self.env.per_tracking_info_dim:]
for i in range(self.env.num_future_data):
delta_x, delta_y, delta_phi = future_path[i * self.env.per_tracking_info_dim:
(i + 1) * self.env.per_tracking_info_dim]
path_x, path_y, path_phi = ego_x + delta_x, ego_y + delta_y, ego_phi - delta_phi
plt.plot(path_x, path_y, 'g.')
plot_phi_line(path_x, path_y, path_phi, 'g')
delta_, _, _ = tracking_info[:3]
indexs, points = self.env.ref_path.find_closest_point(np.array([ego_x], np.float32), np.array([ego_y], np.float32))
path_x, path_y, path_phi = points[0][0], points[1][0], points[2][0]
# plt.plot(path_x, path_y, 'g.')
delta_x, delta_y, delta_phi = ego_x - path_x, ego_y - path_y, ego_phi - path_phi
# plot real time traj
try:
color = ['blue', 'coral', 'darkcyan']
for i, item in enumerate(traj_list):
if i == path_index:
plt.plot(item.path[0], item.path[1], color=color[i])
else:
plt.plot(item.path[0], item.path[1], color=color[i], alpha=0.3)
indexs, points = item.find_closest_point(np.array([ego_x], np.float32), np.array([ego_y], np.float32))
path_x, path_y, path_phi = points[0][0], points[1][0], points[2][0]
plt.plot(path_x, path_y, color=color[i])
except Exception:
pass
# text
# text_x, text_y_start = -120, 60
# ge = iter(range(0, 1000, 4))
# plt.text(text_x, text_y_start - next(ge), 'ego_x: {:.2f}m'.format(ego_x))
# plt.text(text_x, text_y_start - next(ge), 'ego_y: {:.2f}m'.format(ego_y))
# plt.text(text_x, text_y_start - next(ge), 'path_x: {:.2f}m'.format(path_x))
# plt.text(text_x, text_y_start - next(ge), 'path_y: {:.2f}m'.format(path_y))
# plt.text(text_x, text_y_start - next(ge), 'delta_: {:.2f}m'.format(delta_))
# plt.text(text_x, text_y_start - next(ge), 'delta_x: {:.2f}m'.format(delta_x))
# plt.text(text_x, text_y_start - next(ge), 'delta_y: {:.2f}m'.format(delta_y))
# plt.text(text_x, text_y_start - next(ge), r'ego_phi: ${:.2f}\degree$'.format(ego_phi))
# plt.text(text_x, text_y_start - next(ge), r'path_phi: ${:.2f}\degree$'.format(path_phi))
# plt.text(text_x, text_y_start - next(ge), r'delta_phi: ${:.2f}\degree$'.format(delta_phi))
# plt.text(text_x, text_y_start - next(ge), 'v_x: {:.2f}m/s'.format(ego_v_x))
# plt.text(text_x, text_y_start - next(ge), 'exp_v: {:.2f}m/s'.format(self.env.exp_v))
# plt.text(text_x, text_y_start - next(ge), 'v_y: {:.2f}m/s'.format(ego_v_y))
# plt.text(text_x, text_y_start - next(ge), 'yaw_rate: {:.2f}rad/s'.format(ego_r))
# plt.text(text_x, text_y_start - next(ge), 'yaw_rate bound: [{:.2f}, {:.2f}]'.format(-r_bound, r_bound))
#
# plt.text(text_x, text_y_start - next(ge), r'$\alpha_f$: {:.2f} rad'.format(ego_alpha_f))
# plt.text(text_x, text_y_start - next(ge), r'$\alpha_f$ bound: [{:.2f}, {:.2f}] '.format(-alpha_f_bound,
# alpha_f_bound))
# plt.text(text_x, text_y_start - next(ge), r'$\alpha_r$: {:.2f} rad'.format(ego_alpha_r))
# plt.text(text_x, text_y_start - next(ge), r'$\alpha_r$ bound: [{:.2f}, {:.2f}] '.format(-alpha_r_bound,
# alpha_r_bound))
# if self.env.action is not None:
# steer, a_x = self.env.action[0], self.env.action[1]
# plt.text(text_x, text_y_start - next(ge),
# r'steer: {:.2f}rad (${:.2f}\degree$)'.format(steer, steer * 180 / np.pi))
# plt.text(text_x, text_y_start - next(ge), 'a_x: {:.2f}m/s^2'.format(a_x))
#
# text_x, text_y_start = 70, 60
# ge = iter(range(0, 1000, 4))
#
# # done info
# plt.text(text_x, text_y_start - next(ge), 'done info: {}'.format(self.env.done_type))
#
# # reward info
# if self.env.reward_info is not None:
# for key, val in self.env.reward_info.items():
# plt.text(text_x, text_y_start - next(ge), '{}: {:.4f}'.format(key, val))
#
# # indicator for trajectory selection
# text_x, text_y_start = -18, -70
# ge = iter(range(0, 1000, 6))
# if path_values is not None:
# for i, value in enumerate(path_values):
# if i == path_index:
# plt.text(text_x, text_y_start - next(ge), 'Path reward={:.4f}'.format(value[0]), fontsize=14,
# color=color[i], fontstyle='italic')
# else:
# plt.text(text_x, text_y_start - next(ge), 'Path reward={:.4f}'.format(value[0]), fontsize=12,
# color=color[i], fontstyle='italic')
plt.show()
plt.pause(0.001)
if self.logdir is not None:
plt.savefig(self.logdir + '/episode{}'.format(self.episode_counter) + '/step{:03d}.png'.format(self.step_counter))
SummaryPath = '%s/runs-TTQ/TheshFactor-%.3f-Optimizer-%s' \
%(save_root, thresh_factor, optimizer_type)
if args.exp_spec is not '':
SummaryPath += ('-' + args.exp_spec)
print('Save to %s' % SummaryPath)
if os.path.exists(SummaryPath):
print('Record exist, remove')
input()
shutil.rmtree(SummaryPath)
os.makedirs(SummaryPath)
else:
os.makedirs(SummaryPath)
recorder = Recorder(SummaryPath=SummaryPath, dataset_name=dataset_name)
conv1_pos_file = open('%s/conv1_pos.txt' % SummaryPath, 'w+')
conv1_neg_file = open('%s/conv1_neg.txt' % SummaryPath, 'w+')
conv1_prune_file = open('%s/conv1_prune.txt' % SummaryPath, 'w+')
for epoch in range(MAX_EPOCH):
if recorder.stop:
break
net.train()
end = time.time()
recorder.reset_performance()
for batch_idx, (inputs, targets) in enumerate(train_loader):
def plot_and_save_ith_episode_data(logdir, i):
recorder = Recorder()
recorder.load(logdir)
save_dir = logdir + '/episode{}/figs'.format(i)
os.makedirs(save_dir, exist_ok=True)
recorder.plot_and_save_ith_episode_curves(i, save_dir, True)
class HierarchicalDecision(object):
def __init__(self, task, logdir=None):
self.task = task
if self.task == 'left':
self.policy = LoadPolicy('../utils/models/left', 100000)
elif self.task == 'right':
self.policy = LoadPolicy('../utils/models/right', 145000)
elif self.task == 'straight':
self.policy = LoadPolicy('../utils/models/straight', 95000)
self.env = CrossroadEnd2end(training_task=self.task)
self.model = EnvironmentModel(self.task, mode='selecting')
self.recorder = Recorder()
self.episode_counter = -1
self.step_counter = -1
self.obs = None
self.stg = None
self.step_timer = TimerStat()
self.ss_timer = TimerStat()
self.logdir = logdir
self.fig = plt.figure(figsize=(8, 8))
plt.ion()
self.hist_posi = []
self.reset()
def reset(self, ):
self.obs = self.env.reset()
self.stg = MultiPathGenerator()
self.recorder.reset()
self.hist_posi = []
if self.logdir is not None:
self.episode_counter += 1
os.makedirs(self.logdir +
'/episode{}'.format(self.episode_counter))
self.step_counter = -1
self.recorder.save(self.logdir)
return self.obs
def safe_shield(self, real_obs, traj):
action_bound = 1.0
action_safe_set = [[[0, -action_bound]]]
real_obs = tf.convert_to_tensor(real_obs[np.newaxis, :])
obs = real_obs
self.model.add_traj(obs, traj)
total_punishment = 0.0
for step in range(3):
action = self.policy.run(obs)
_, _, _, _, veh2veh4real, _ = self.model.rollout_out(action)
total_punishment += veh2veh4real
if total_punishment != 0:
print('original action will cause collision within three steps!!!')
for safe_action in action_safe_set:
obs = real_obs
total_punishment = 0
# for step in range(1):
# obs, veh2veh4real = self.model.safety_calculation(obs, safe_action)
# total_punishment += veh2veh4real
# if veh2veh4real != 0: # collide
# break
# if total_punishment == 0:
# print('found the safe action', safe_action)
# safe_action = np.array(safe_action)
# break
# else:
# print('still collide')
# safe_action = self.policy.run(real_obs).numpy().squeeze(0)
return np.array(safe_action).squeeze(0)
else:
safe_action = self.policy.run(real_obs).numpy()[0]
return safe_action
def step(self):
self.step_counter += 1
with self.step_timer:
path_list = self.stg.generate_path(self.task)
obs_list = []
# select optimal path
for path in path_list:
self.env.set_traj(path)
obs_list.append(self.env._get_obs())
all_obs = tf.stack(obs_list, axis=0)
path_values = self.policy.values(all_obs).numpy().squeeze()
path_index = int(np.argmax(path_values[:, 0]))
self.env.set_traj(path_list[path_index])
self.obs_real = obs_list[path_index]
# obtain safe action
# with self.ss_timer:
# safe_action = self.safe_shield(self.obs_real, path_list[path_index])
safe_action = self.policy.run(self.obs_real).numpy()
print('ALL TIME:', self.step_timer.mean)
# deal with red light temporally
# if self.env.v_light != 0 and -25 > self.env.ego_dynamics['y'] > -35 and self.env.training_task != 'right':
# scaled_steer = 0.
# if self.env.ego_dynamics['v_x'] == 0.0:
# scaled_a_x = 0.33
# else:
# scaled_a_x = np.random.uniform(-0.6, -0.4)
# safe_action = np.array([scaled_steer, scaled_a_x], dtype=np.float32)
self.render(path_list, path_values, path_index)
self.recorder.record(self.obs_real, safe_action, self.step_timer.mean,
path_index, path_values, self.ss_timer.mean)
self.obs, r, done, info = self.env.step(safe_action)
return done
def render(self, traj_list, path_values, path_index):
square_length = CROSSROAD_SIZE
extension = 40
lane_width = LANE_WIDTH
light_line_width = 3
dotted_line_style = '--'
solid_line_style = '-'
plt.cla()
ax = plt.axes([-0.05, -0.05, 1.1, 1.1])
for ax in self.fig.get_axes():
ax.axis('off')
ax.axis("equal")
# ----------arrow--------------
plt.arrow(lane_width / 2, -square_length / 2 - 10, 0, 5, color='b')
plt.arrow(lane_width / 2,
-square_length / 2 - 10 + 5,
-0.5,
0,
color='b',
head_width=1)
plt.arrow(lane_width * 1.5,
-square_length / 2 - 10,
0,
4,
color='b',
head_width=1)
plt.arrow(lane_width * 2.5, -square_length / 2 - 10, 0, 5, color='b')
plt.arrow(lane_width * 2.5,
-square_length / 2 - 10 + 5,
0.5,
0,
color='b',
head_width=1)
# ----------horizon--------------
plt.plot([-square_length / 2 - extension, -square_length / 2],
[0.3, 0.3],
color='orange')
plt.plot([-square_length / 2 - extension, -square_length / 2],
[-0.3, -0.3],
color='orange')
plt.plot([square_length / 2 + extension, square_length / 2],
[0.3, 0.3],
color='orange')
plt.plot([square_length / 2 + extension, square_length / 2],
[-0.3, -0.3],
color='orange')
#
for i in range(1, LANE_NUMBER + 1):
linestyle = dotted_line_style if i < LANE_NUMBER else solid_line_style
linewidth = 1 if i < LANE_NUMBER else 2
plt.plot([-square_length / 2 - extension, -square_length / 2],
[i * lane_width, i * lane_width],
linestyle=linestyle,
color='black',
linewidth=linewidth)
plt.plot([square_length / 2 + extension, square_length / 2],
[i * lane_width, i * lane_width],
linestyle=linestyle,
color='black',
linewidth=linewidth)
plt.plot([-square_length / 2 - extension, -square_length / 2],
[-i * lane_width, -i * lane_width],
linestyle=linestyle,
color='black',
linewidth=linewidth)
plt.plot([square_length / 2 + extension, square_length / 2],
[-i * lane_width, -i * lane_width],
linestyle=linestyle,
color='black',
linewidth=linewidth)
# ----------vertical----------------
plt.plot([0.3, 0.3],
[-square_length / 2 - extension, -square_length / 2],
color='orange')
plt.plot([-0.3, -0.3],
[-square_length / 2 - extension, -square_length / 2],
color='orange')
plt.plot([0.3, 0.3],
[square_length / 2 + extension, square_length / 2],
color='orange')
plt.plot([-0.3, -0.3],
[square_length / 2 + extension, square_length / 2],
color='orange')
#
for i in range(1, LANE_NUMBER + 1):
linestyle = dotted_line_style if i < LANE_NUMBER else solid_line_style
linewidth = 1 if i < LANE_NUMBER else 2
plt.plot([i * lane_width, i * lane_width],
[-square_length / 2 - extension, -square_length / 2],
linestyle=linestyle,
color='black',
linewidth=linewidth)
plt.plot([i * lane_width, i * lane_width],
[square_length / 2 + extension, square_length / 2],
linestyle=linestyle,
color='black',
linewidth=linewidth)
plt.plot([-i * lane_width, -i * lane_width],
[-square_length / 2 - extension, -square_length / 2],
linestyle=linestyle,
color='black',
linewidth=linewidth)
plt.plot([-i * lane_width, -i * lane_width],
[square_length / 2 + extension, square_length / 2],
linestyle=linestyle,
color='black',
linewidth=linewidth)
v_light = self.env.v_light
if v_light == 0:
v_color, h_color = 'green', 'red'
elif v_light == 1:
v_color, h_color = 'orange', 'red'
elif v_light == 2:
v_color, h_color = 'red', 'green'
else:
v_color, h_color = 'red', 'orange'
plt.plot([0, (LANE_NUMBER - 1) * lane_width],
[-square_length / 2, -square_length / 2],
color=v_color,
linewidth=light_line_width)
plt.plot([(LANE_NUMBER - 1) * lane_width, LANE_NUMBER * lane_width],
[-square_length / 2, -square_length / 2],
color='green',
linewidth=light_line_width)
plt.plot([-LANE_NUMBER * lane_width, -(LANE_NUMBER - 1) * lane_width],
[square_length / 2, square_length / 2],
color='green',
linewidth=light_line_width)
plt.plot([-(LANE_NUMBER - 1) * lane_width, 0],
[square_length / 2, square_length / 2],
color=v_color,
linewidth=light_line_width)
plt.plot([-square_length / 2, -square_length / 2],
[0, -(LANE_NUMBER - 1) * lane_width],
color=h_color,
linewidth=light_line_width)
plt.plot([-square_length / 2, -square_length / 2],
[-(LANE_NUMBER - 1) * lane_width, -LANE_NUMBER * lane_width],
color='green',
linewidth=light_line_width)
plt.plot([square_length / 2, square_length / 2],
[(LANE_NUMBER - 1) * lane_width, 0],
color=h_color,
linewidth=light_line_width)
plt.plot([square_length / 2, square_length / 2],
[LANE_NUMBER * lane_width, (LANE_NUMBER - 1) * lane_width],
color='green',
linewidth=light_line_width)
# ----------Oblique--------------
plt.plot([LANE_NUMBER * lane_width, square_length / 2],
[-square_length / 2, -LANE_NUMBER * lane_width],
color='black',
linewidth=2)
plt.plot([LANE_NUMBER * lane_width, square_length / 2],
[square_length / 2, LANE_NUMBER * lane_width],
color='black',
linewidth=2)
plt.plot([-LANE_NUMBER * lane_width, -square_length / 2],
[-square_length / 2, -LANE_NUMBER * lane_width],
color='black',
linewidth=2)
plt.plot([-LANE_NUMBER * lane_width, -square_length / 2],
[square_length / 2, LANE_NUMBER * lane_width],
color='black',
linewidth=2)
def is_in_plot_area(x, y, tolerance=5):
if -square_length / 2 - extension + tolerance < x < square_length / 2 + extension - tolerance and \
-square_length / 2 - extension + tolerance < y < square_length / 2 + extension - tolerance:
return True
else:
return False
def draw_rotate_rec(x, y, a, l, w, c):
bottom_left_x, bottom_left_y, _ = rotate_coordination(
-l / 2, w / 2, 0, -a)
ax.add_patch(
plt.Rectangle((x + bottom_left_x, y + bottom_left_y),
w,
l,
edgecolor=c,
facecolor='white',
angle=-(90 - a),
zorder=50))
def plot_phi_line(x, y, phi, color):
line_length = 3
x_forw, y_forw = x + line_length * cos(phi * pi / 180.), \
y + line_length * sin(phi * pi / 180.)
plt.plot([x, x_forw], [y, y_forw], color=color, linewidth=0.5)
# plot cars
for veh in self.env.all_vehicles:
veh_x = veh['x']
veh_y = veh['y']
veh_phi = veh['phi']
veh_l = veh['l']
veh_w = veh['w']
if is_in_plot_area(veh_x, veh_y):
plot_phi_line(veh_x, veh_y, veh_phi, 'black')
draw_rotate_rec(veh_x, veh_y, veh_phi, veh_l, veh_w, 'black')
# plot_interested vehs
# for mode, num in self.veh_mode_dict.items():
# for i in range(num):
# veh = self.interested_vehs[mode][i]
# veh_x = veh['x']
# veh_y = veh['y']
# veh_phi = veh['phi']
# veh_l = veh['l']
# veh_w = veh['w']
# task2color = {'left': 'b', 'straight': 'c', 'right': 'm'}
#
# if is_in_plot_area(veh_x, veh_y):
# plot_phi_line(veh_x, veh_y, veh_phi, 'black')
# task = MODE2TASK[mode]
# color = task2color[task]
# draw_rotate_rec(veh_x, veh_y, veh_phi, veh_l, veh_w, color, linestyle=':')
ego_v_x = self.env.ego_dynamics['v_x']
ego_v_y = self.env.ego_dynamics['v_y']
ego_r = self.env.ego_dynamics['r']
ego_x = self.env.ego_dynamics['x']
ego_y = self.env.ego_dynamics['y']
ego_phi = self.env.ego_dynamics['phi']
ego_l = self.env.ego_dynamics['l']
ego_w = self.env.ego_dynamics['w']
ego_alpha_f = self.env.ego_dynamics['alpha_f']
ego_alpha_r = self.env.ego_dynamics['alpha_r']
alpha_f_bound = self.env.ego_dynamics['alpha_f_bound']
alpha_r_bound = self.env.ego_dynamics['alpha_r_bound']
r_bound = self.env.ego_dynamics['r_bound']
plot_phi_line(ego_x, ego_y, ego_phi, 'fuchsia')
draw_rotate_rec(ego_x, ego_y, ego_phi, ego_l, ego_w, 'fuchsia')
self.hist_posi.append((ego_x, ego_y))
# plot history data
for hist_x, hist_y in self.hist_posi:
plt.scatter(hist_x, hist_y, color='fuchsia', alpha=0.1)
# plot future data
tracking_info = self.obs[self.env.ego_info_dim:self.env.ego_info_dim +
self.env.per_tracking_info_dim *
(self.env.num_future_data + 1)]
future_path = tracking_info[self.env.per_tracking_info_dim:]
for i in range(self.env.num_future_data):
delta_x, delta_y, delta_phi = future_path[
i * self.env.per_tracking_info_dim:(i + 1) *
self.env.per_tracking_info_dim]
path_x, path_y, path_phi = ego_x + delta_x, ego_y + delta_y, ego_phi - delta_phi
plt.plot(path_x, path_y, 'g.')
plot_phi_line(path_x, path_y, path_phi, 'g')
delta_, _, _ = tracking_info[:3]
indexs, points = self.env.ref_path.find_closest_point(
np.array([ego_x], np.float32), np.array([ego_y], np.float32))
path_x, path_y, path_phi = points[0][0], points[1][0], points[2][0]
# plt.plot(path_x, path_y, 'g.')
delta_x, delta_y, delta_phi = ego_x - path_x, ego_y - path_y, ego_phi - path_phi
# plot real time traj
try:
color = ['blue', 'coral', 'darkcyan']
for i, item in enumerate(traj_list):
if i == path_index:
plt.plot(item.path[0], item.path[1], color=color[i])
else:
plt.plot(item.path[0],
item.path[1],
color=color[i],
alpha=0.3)
indexs, points = item.find_closest_point(
np.array([ego_x], np.float32), np.array([ego_y],
np.float32))
path_x, path_y, path_phi = points[0][0], points[1][0], points[
2][0]
plt.plot(path_x, path_y, color=color[i])
except Exception:
pass
# text
# text_x, text_y_start = -120, 60
# ge = iter(range(0, 1000, 4))
# plt.text(text_x, text_y_start - next(ge), 'ego_x: {:.2f}m'.format(ego_x))
# plt.text(text_x, text_y_start - next(ge), 'ego_y: {:.2f}m'.format(ego_y))
# plt.text(text_x, text_y_start - next(ge), 'path_x: {:.2f}m'.format(path_x))
# plt.text(text_x, text_y_start - next(ge), 'path_y: {:.2f}m'.format(path_y))
# plt.text(text_x, text_y_start - next(ge), 'delta_: {:.2f}m'.format(delta_))
# plt.text(text_x, text_y_start - next(ge), 'delta_x: {:.2f}m'.format(delta_x))
# plt.text(text_x, text_y_start - next(ge), 'delta_y: {:.2f}m'.format(delta_y))
# plt.text(text_x, text_y_start - next(ge), r'ego_phi: ${:.2f}\degree$'.format(ego_phi))
# plt.text(text_x, text_y_start - next(ge), r'path_phi: ${:.2f}\degree$'.format(path_phi))
# plt.text(text_x, text_y_start - next(ge), r'delta_phi: ${:.2f}\degree$'.format(delta_phi))
# plt.text(text_x, text_y_start - next(ge), 'v_x: {:.2f}m/s'.format(ego_v_x))
# plt.text(text_x, text_y_start - next(ge), 'exp_v: {:.2f}m/s'.format(self.env.exp_v))
# plt.text(text_x, text_y_start - next(ge), 'v_y: {:.2f}m/s'.format(ego_v_y))
# plt.text(text_x, text_y_start - next(ge), 'yaw_rate: {:.2f}rad/s'.format(ego_r))
# plt.text(text_x, text_y_start - next(ge), 'yaw_rate bound: [{:.2f}, {:.2f}]'.format(-r_bound, r_bound))
#
# plt.text(text_x, text_y_start - next(ge), r'$\alpha_f$: {:.2f} rad'.format(ego_alpha_f))
# plt.text(text_x, text_y_start - next(ge), r'$\alpha_f$ bound: [{:.2f}, {:.2f}] '.format(-alpha_f_bound,
# alpha_f_bound))
# plt.text(text_x, text_y_start - next(ge), r'$\alpha_r$: {:.2f} rad'.format(ego_alpha_r))
# plt.text(text_x, text_y_start - next(ge), r'$\alpha_r$ bound: [{:.2f}, {:.2f}] '.format(-alpha_r_bound,
# alpha_r_bound))
# if self.env.action is not None:
# steer, a_x = self.env.action[0], self.env.action[1]
# plt.text(text_x, text_y_start - next(ge),
# r'steer: {:.2f}rad (${:.2f}\degree$)'.format(steer, steer * 180 / np.pi))
# plt.text(text_x, text_y_start - next(ge), 'a_x: {:.2f}m/s^2'.format(a_x))
#
# text_x, text_y_start = 70, 60
# ge = iter(range(0, 1000, 4))
#
# # done info
# plt.text(text_x, text_y_start - next(ge), 'done info: {}'.format(self.env.done_type))
#
# # reward info
# if self.env.reward_info is not None:
# for key, val in self.env.reward_info.items():
# plt.text(text_x, text_y_start - next(ge), '{}: {:.4f}'.format(key, val))
#
# # indicator for trajectory selection
# text_x, text_y_start = -18, -70
# ge = iter(range(0, 1000, 6))
# if path_values is not None:
# for i, value in enumerate(path_values):
# if i == path_index:
# plt.text(text_x, text_y_start - next(ge), 'Path reward={:.4f}'.format(value[0]), fontsize=14,
# color=color[i], fontstyle='italic')
# else:
# plt.text(text_x, text_y_start - next(ge), 'Path reward={:.4f}'.format(value[0]), fontsize=12,
# color=color[i], fontstyle='italic')
plt.show()
plt.pause(0.001)
if self.logdir is not None:
plt.savefig(self.logdir +
'/episode{}'.format(self.episode_counter) +
'/step{}.pdf'.format(self.step_counter))