def drive(args):
with open(args.conf.strip(), 'r') as f:
config = json.load(f)
modules = load_modules(config['models'])
model = load_agent('imitation_learning_agent').load('lstm-2.h5')
agent = ControlAgent(model, modules)
experiment_suite = CoRL2017('Town01')
run_driving_benchmark(agent, experiment_suite)
# download dataset if it doesn't exist
if not os.path.exists(agent_params.memory.load_memory_from_file_path):
screen.log_title("The CARLA dataset is not present in the following path: {}"
.format(agent_params.memory.load_memory_from_file_path))
result = screen.ask_yes_no("Do you want to download it now?")
if result:
create_dataset(None, "./datasets/carla_train_set_replay_buffer.p")
else:
screen.error("Please update the path to the CARLA dataset in the CARLA_CIL preset", crash=True)
###############
# Environment #
###############
env_params = CarlaEnvironmentParameters()
env_params.level = 'town1'
env_params.cameras = ['CameraRGB']
env_params.camera_height = 600
env_params.camera_width = 800
env_params.separate_actions_for_throttle_and_brake = True
env_params.allow_braking = True
env_params.quality = CarlaEnvironmentParameters.Quality.EPIC
env_params.experiment_suite = CoRL2017('Town01')
vis_params = VisualizationParameters()
vis_params.video_dump_methods = [SelectedPhaseOnlyDumpMethod(RunPhase.TEST)]
vis_params.dump_mp4 = True
graph_manager = BasicRLGraphManager(agent_params=agent_params, env_params=env_params,
schedule_params=schedule_params, vis_params=vis_params)
'--log_name',
metavar='T',
default='test',
help='The name of the log file to be created by the scripts')
argparser.add_argument(
'--avoid-stopping',
default=True,
action='store_false',
help=' Uses the speed prediction branch to avoid unwanted agent stops')
argparser.add_argument(
'--continue-experiment',
action='store_true',
help='If you want to continue the experiment with the given log name')
args = argparser.parse_args()
log_level = logging.DEBUG if args.debug else logging.INFO
logging.basicConfig(format='%(levelname)s: %(message)s', level=log_level)
logging.info('listening to server %s:%s', args.host, args.port)
agent = ImitationLearning(args.city_name, args.avoid_stopping)
if args.test_name == 'CORL2017':
corl = CoRL2017(args.city_name)
else:
corl = trainingData(args.city_name)
# Now actually run the driving_benchmark
run_driving_benchmark(agent, corl, args.city_name, args.log_name,
args.continue_experiment, args.host, args.port)
args = argparser.parse_args()
if args.debug:
log_level = logging.DEBUG
elif args.verbose:
log_level = logging.INFO
else:
log_level = logging.WARNING
logging.basicConfig(format='%(levelname)s: %(message)s', level=log_level)
logging.info('listening to server %s:%s', args.host, args.port)
# We instantiate a forward agent, a simple policy that just set
# acceleration as 0.9 and steering as zero
agent = L5Agent()
# We instantiate an experiment suite. Basically a set of experiments
# that are going to be evaluated on this benchmark.
if args.corl_2017:
experiment_suite = CoRL2017(args.city_name)
else:
print (' WARNING: running the basic driving benchmark, to run for CoRL 2017'
' experiment suites, you should run'
' python driving_benchmark_example.py --corl-2017')
experiment_suite = BasicExperimentSuite(args.city_name)
# Now actually run the driving_benchmark
run_driving_benchmark(agent, experiment_suite, args.city_name,
args.log_name, args.continue_experiment,
args.host, args.port)
help='The name of the model for writing in the reports')
argparser.add_argument('-p',
'--path',
metavar='P',
default='test',
help='Path to all log files')
args = argparser.parse_args()
log_level = logging.DEBUG if args.debug else logging.INFO
logging.basicConfig(format='%(levelname)s: %(message)s', level=log_level)
logging.info('sarting the calculations %s',
"0") #TODO: add time instead on zero
experiment_suite = CoRL2017("Town01")
metrics_object = Metrics(experiment_suite.metrics_parameters,
experiment_suite.dynamic_tasks)
# Improve readability by adding a weather dictionary
weather_name_dict = {
1: 'Clear Noon',
3: 'After Rain Noon',
6: 'Heavy Rain Noon',
8: 'Clear Sunset',
4: 'Cloudy After Rain',
14: 'Soft Rain Sunset'
}
# names for all the test logs
def __init__(self,
num_speedup_steps=30,
require_explicit_reset=True,
is_render_enabled=False,
early_termination_enabled=False,
run_offscreen=False,
save_screens=False,
port=2000,
gpu=0,
discrete_control=True,
kill_when_connection_lost=True,
city_name="Town01",
channel_last=True):
EnvironmentWrapper.__init__(self, is_render_enabled, save_screens)
print("port:", port)
self.episode_max_time = 1000000
self.allow_braking = True
self.log_path = os.path.join(DEFAULT_CARLA_LOG_DIR, "CarlaLogs.txt")
self.num_speedup_steps = num_speedup_steps
self.is_game_ready_for_input = False
self.run_offscreen = run_offscreen
self.kill_when_connection_lost = kill_when_connection_lost
# server configuration
self.port = port
self.gpu = gpu
self.host = 'localhost'
self.level = 'town1'
self.map = CarlaLevel().get(self.level)
# experiment = basic_experiment_suite.BasicExperimentSuite(city_name)
experiment = CoRL2017(city_name)
self.experiments = experiment.get_experiments()
self.experiment_type = 0
self.planner = Planner(city_name)
self.car_speed = 0
self.is_game_setup = False # Will be true only when setup_client_and_server() is called, either explicitly, or by reset()
# action space
self.discrete_controls = discrete_control
self.action_space_size = 3
self.action_space_high = np.array([1, 1, 1])
self.action_space_low = np.array([-1, -1, -1])
self.action_space_abs_range = np.maximum(
np.abs(self.action_space_low), np.abs(self.action_space_high))
self.steering_strength = 0.35
self.gas_strength = 1.0
self.brake_strength = 0.6
self.actions = {
0: [0., 0.],
1: [0., -self.steering_strength],
2: [0., self.steering_strength],
3: [self.gas_strength - 0.15, 0.],
4: [-self.brake_strength, 0],
5: [self.gas_strength - 0.3, -self.steering_strength],
6: [self.gas_strength - 0.3, self.steering_strength],
7: [-self.brake_strength, -self.steering_strength],
8: [-self.brake_strength, self.steering_strength]
}
self.actions_description = [
'NO-OP', 'TURN_LEFT', 'TURN_RIGHT', 'GAS', 'BRAKE',
'GAS_AND_TURN_LEFT', 'GAS_AND_TURN_RIGHT', 'BRAKE_AND_TURN_LEFT',
'BRAKE_AND_TURN_RIGHT'
]
if discrete_control:
self.action_space = Discrete(len(self.actions))
else:
self.action_space = Box(low=self.action_space_low,
high=self.action_space_high)
self.observation_space = Box(low=-np.inf,
high=np.inf,
shape=[88, 200, 3])
# measurements
self.measurements_size = (1, )
self.pre_image = None
self.first_debug = True
self.channel_last = channel_last
