def __init__(self,
action_wrapper,
observer,
evaluator,
step_time,
viewer,
scenario_generator=None,
render=False):
Runtime.__init__(self,
step_time=step_time,
viewer=viewer,
scenario_generator=scenario_generator,
render=render)
self._action_wrapper = action_wrapper
self._observer = observer
self._evaluator = evaluator
def test_python_model(self):
param_server = ParameterServer(
filename="modules/runtime/tests/data/deterministic_scenario.json")
scenario_generation = DeterministicScenarioGeneration(num_scenarios=3,
random_seed=0,
params=param_server)
viewer = MPViewer(params=param_server,
follow_agent_id=False,
use_world_bounds=True)
env = Runtime(0.2,
viewer,
scenario_generation,
render=True)
behavior_model = DummyBehaviorModel(param_server)
env.reset()
env._world.get_agent(0).behavior_model = behavior_model
env._world.get_agent(0).behavior_model.clone()
np.testing.assert_array_equal(
env._world.get_agent(0).behavior_model.plan(0.2, env._world)[1],
np.array([0.2, 5111.626, 5106.8305 + 0.2, 1.5, 10]))
env.reset()
env._world.get_agent(0).behavior_model = behavior_model
env._world.step(0.2)
np.testing.assert_array_equal(
env._world.get_agent(0).state,
np.array([0.2, 5111.626, 5106.8305 + 0.2, 1.5, 10], dtype=np.float32))
env._world.step(0.2)
np.testing.assert_array_equal(
env._world.get_agent(0).state,
np.array([0.4, 5111.626, 5106.8305 + 0.4, 1.5, 10], dtype=np.float32))
env._world.step(0.2)
np.testing.assert_array_equal(
env._world.get_agent(0).state,
np.array([0.6, 5111.626, 5106.8305 + 0.6, 1.5, 10], dtype=np.float32))
print("History:", env._world.get_agent(0).history)
# environment loop
env.reset()
for i in range(0, 7):
env.step()
def test_runtime(self):
"""Asserts the runtime to make sure the basic
functionality is given by the current state of BARK.
"""
param_server = ParameterServer(
filename="tests/data/deterministic_scenario_test.json")
scenario_generation = DeterministicScenarioGeneration(num_scenarios=3,
random_seed=0,
params=param_server)
param_server["Visualization"]["Agents"]["DrawRoute",
"Draw the local map of each agent",
True]
viewer = MPViewer(params=param_server,
use_world_bounds=True)
env = Runtime(0.2,
viewer,
scenario_generation,
render=True)
env.reset()
agent_ids = []
agent_states = []
centers_of_driving_corridor= []
for key, agent in env._world.agents.items():
agent_ids.append(agent.id)
agent_states.append(agent.state)
for i in range(0, 5):
print("Scenario {}:".format(
str(env._scenario_generator._current_scenario_idx)))
# assert scenario ids
self.assertEqual(env._scenario_generator._current_scenario_idx, (i % 3) + 1)
for _ in range(0, 35):
# assert ids
states_before = []
for i, (key, agent) in enumerate(env._world.agents.items()):
self.assertEqual(key, agent.id)
self.assertEqual(agent_ids[i], agent.id)
states_before.append(agent.state)
# TODO(@hart): why does this not work
print(key, agent.goal_definition.goal_shape)
env.step()
# assert state has been changed by the step() function
for i, (key, agent) in enumerate(env._world.agents.items()):
np.testing.assert_equal(np.any(np.not_equal(states_before[i],
agent.state)), True)
# check whether the reset works
env.reset()
for i, (key, agent) in enumerate(env._world.agents.items()):
self.assertEqual(key, agent.id)
self.assertEqual(agent_ids[i], agent.id)
np.testing.assert_array_equal(agent_states[i], agent.state)
def test_python_model(self):
param_server = ParameterServer(
filename="modules/runtime/tests/data/deterministic_scenario.json")
scenario_generation = DeterministicScenarioGeneration(num_scenarios=3,
random_seed=0,
params=param_server)
viewer = MPViewer(params=param_server,
follow_agent_id=False,
use_world_bounds=True)
env = Runtime(0.2,
viewer,
scenario_generation,
render=True)
single_track_model = SingleTrackModel(param_server)
behavior_model = DynamicBehaviorModel(single_track_model, param_server)
env.reset()
env._world.get_agent(0).behavior_model = behavior_model
env._world.get_agent(0).behavior_model.clone()
env.reset()
env._world.get_agent(0).behavior_model = behavior_model
env._world.get_agent(0).behavior_model.set_last_action(np.array([1., 2.]))
print(env._world.get_agent(0).behavior_model.get_last_action())
env._world.step(0.2)
print(env._world.get_agent(0).behavior_model.get_last_action())
def test_bark_agent_in_runtime(self):
# check whether the bark agent really does work
params = ParameterServer(
filename="configurations/sac_highway/config.json")
configuration = SACHighwayConfiguration(params)
scenario_generator = configuration._scenario_generator
viewer = MPViewer(params=params,
x_range=[-50, 50],
y_range=[-50, 50],
follow_agent_id=100,
screen_dims=[500, 500],
use_world_bounds=False)
env = Runtime(0.2, viewer, scenario_generator, render=True)
env.reset()
dynamic_model = env._world.agents[
env._scenario._eval_agent_ids[0]].dynamic_model
bark_agent = BARKMLBehaviorModel(configuration, dynamic_model,
env._scenario._eval_agent_ids)
bark_agent.set_last_action(np.array([1., 0.]))
env._world.get_agent(
env._scenario._eval_agent_ids[0]).behavior_model = bark_agent
for _ in range(0, 35):
env.step()
def test_visualization(self):
param_server = ParameterServer(
filename="modules/runtime/tests/data/deterministic_scenario.json")
scenario_generation = DeterministicScenarioGeneration(
num_scenarios=3, random_seed=0, params=param_server)
viewer = MPViewer(params=param_server,
follow_agent_id=False,
use_world_bounds=True)
env = Runtime(0.2, viewer, scenario_generation, render=True)
env.reset()
for _ in range(0, 5):
print("Scenario {}:".format(
str(env._scenario_generator._current_scenario_idx)))
for _ in range(0, 5):
env.step()
env.reset()
params = ParameterServer(filename=base_dir +
"/configurations/highway/config.json")
scenario_generation = params["Scenario"]["Generation"][
"ConfigurableScenarioGeneration"]
map_filename = scenario_generation["MapFilename"]
scenario_generation["MapFilename"] = base_dir + "/" + map_filename
params["BaseDir"] = base_dir
# actual model
ml_config = HighwayConfiguration(params)
ml_behavior = BARKMLBehaviorModel(configuration=ml_config)
scenario_generator = ml_config._scenario_generator
viewer = MPViewer(params=params, use_world_bounds=True)
env = Runtime(0.2, viewer, scenario_generator, render=True)
env.reset()
env._world.agents[
env._scenario._eval_agent_ids[0]].behavior_model = ml_behavior
print(ml_behavior)
for _ in range(0, 50):
env.step()
# to find database files
os.chdir("../benchmark_database/")
dbs = DatabaseSerializer(test_scenarios=4,
test_world_steps=5,
num_serialize_scenarios=10)
dbs.process("database")
# viewer
viewer = MPViewer(params=param_server, use_world_bounds=True)
# viewer = Panda3dViewer(params=param_server,
# x_range=[-40, 40],
# y_range=[-40, 40],
# follow_agent_id=agent3.id)
# World Simulation
sim_step_time = param_server["simulation"]["step_time",
"Step-time used in simulation", 0.2]
sim_real_time_factor = param_server["simulation"][
"real_time_factor", "execution in real-time or faster", 1.]
# viewer = VideoRenderer(renderer=viewer,
# world_step_time=sim_step_time,
# fig_path="/Users/hart/2019/bark/video")
env = Runtime(step_time=0.2,
viewer=viewer,
scenario_generator=scenarios,
render=True)
# run 3 scenarios
for _ in range(0, 3):
env.reset()
# step each scenario 20 times
for step in range(0, 80):
env.step()
time.sleep(sim_step_time / sim_real_time_factor)
# viewer.export_video(filename="/Users/hart/2019/bark/video/intersection", remove_image_dir=True)
scenario_generation = UniformVehicleDistribution(num_scenarios=10,
random_seed=0,
params=param_server)
viewer = MPViewer(params=param_server,
x_range=[5060, 5160],
y_range=[5070, 5150])
sim_step_time = param_server["simulation"]["step_time",
"Step-time used in simulation", 0.2]
sim_real_time_factor = param_server["simulation"][
"real_time_factor", "execution in real-time or faster", 1]
scenario, idx = scenario_generation.get_next_scenario()
# Rendering WITHOUT intermediate steps
video_renderer = VideoRenderer(renderer=viewer, world_step_time=sim_step_time)
env = Runtime(0.2, video_renderer, scenario_generation, render=True)
env.reset()
for _ in range(0, 5):
env.step()
video_renderer.export_video(filename="examples/scenarios/test_video_step")
# Rendering WITH intermediate steps
video_renderer = VideoRenderer(renderer=viewer,
world_step_time=sim_step_time,
render_intermediate_steps=10)
env = Runtime(0.2, video_renderer, scenario_generation, render=True)
env.reset()
for _ in range(0, 5):
env.step()
video_renderer.export_video(