def load(self) -> 'SimulationData':
with open(self.path_json, 'r') as f:
obj = json.loads(f.read())
info = SimulationInfo()
info.__dict__ = obj.get(self.f_info, {})
self.set(SimulationParams(**obj[self.f_params]),
DecalRoad.from_dict(obj[self.f_road]),
[SimulationDataRecord(**r) for r in obj[self.f_records]],
info=info)
return self
def run_sim(street_1: DecalRoad, ai_aggression):
waypoint_goal = BeamNGWaypoint('waypoint_goal', get_node_coords(street_1.nodes[-1]))
maps.beamng_map.generated().write_items(street_1.to_json() + '\n' + waypoint_goal.to_json())
beamng = BeamNGpy('localhost', 64256)
scenario = Scenario('tig', 'tigscenario')
vehicle = Vehicle('ego_vehicle', model='etk800', licence='TIG', color='Red')
sim_data_collector = TrainingDataCollectorAndWriter(vehicle, beamng, street_1)
scenario.add_vehicle(vehicle, pos=get_node_coords(street_1.nodes[0]), rot=get_rotation(street_1))
scenario.make(beamng)
beamng.open()
beamng.set_deterministic()
beamng.load_scenario(scenario)
beamng.pause()
beamng.start_scenario()
vehicle.ai_set_aggression(ai_aggression)
vehicle.ai_drive_in_lane(False)
vehicle.ai_set_speed(25)
vehicle.ai_set_waypoint(waypoint_goal.name)
#vehicle.ai_set_mode("manual")
steps = 5
def start():
for idx in range(1000):
if (idx * 0.05 * steps) > 3.:
sim_data_collector.collect_and_write_current_data()
if sim_data_collector.oob_monitor.is_oob(wrt="center"):
raise ValueError("OOB detected during training")
dist = distance(sim_data_collector.last_state.pos, waypoint_goal.position)
if dist < 15.0:
beamng.resume()
break
# one step is 0.05 seconds (5/100)
beamng.step(steps)
try:
start()
finally:
beamng.close()
try:
start()
finally:
beamng.close()
if __name__ == '__main__':
dataset_size = 10
road_storage = SeedStorage('training_set')
simulation_attempts = 5
for idx in range(1, dataset_size + 1):
path = road_storage.get_path_by_index(idx)
road_generator = RoadGenerator(num_control_nodes=30, seg_length=20).generate()
with open(path, 'w') as f:
f.write(json.dumps(road_generator.to_dict()))
street_1 = DecalRoad('street_1').add_4d_points(road_generator.sample_nodes)
for attempt in range(simulation_attempts):
try:
print("Simulation number =", attempt)
run_sim(street_1=street_1, ai_aggression=0.1)
break
except ValueError:
pass
def setup_road_nodes(self, road_nodes):
self.road_nodes = road_nodes
self.decal_road: DecalRoad = DecalRoad('street_1').add_4d_points(
road_nodes)
self.road_points = RoadPoints().add_middle_nodes(road_nodes)
def initial_population_generator(path, config, problem):
all_roads = [
filename
for filename in glob.glob(str(path) + "\*.json", recursive=True)
]
type = config.Feature_Combination
shuffle(all_roads)
roads = all_roads
original_set = list()
individuals = []
popsize = config.POPSIZE
for road in roads:
with open(road) as json_file:
data = json.load(json_file)
sample_nodes = data["road"]["nodes"]
for node in sample_nodes:
node[2] = -28.0
sample_nodes = np.array(sample_nodes)
records = data["records"]
bbox_size = (-250.0, 0.0, 250.0, 500.0)
road_bbox = RoadBoundingBox(bbox_size)
member = BeamNGMember(data["control_nodes"],
[tuple(t) for t in sample_nodes], 20, road_bbox)
member.config = config
member.problem = problem
simulation_id = time.strftime('%Y-%m-%d--%H-%M-%S', time.localtime())
sim_name = member.config.simulation_name.replace(
'$(id)', simulation_id)
simulation_data = SimulationData(sim_name)
states = []
for record in records:
state = VehicleState(timer=record["timer"],
pos=record["pos"],
dir=record["dir"],
vel=record["vel"],
steering=record["steering"],
steering_input=record["steering_input"],
brake=record["brake"],
brake_input=record["brake_input"],
throttle=record["throttle"],
throttle_input=record["throttle_input"],
wheelspeed=record["wheelspeed"],
vel_kmh=record["vel_kmh"])
sim_data_record = SimulationDataRecord(
**state._asdict(),
is_oob=record["is_oob"],
oob_counter=record["oob_counter"],
max_oob_percentage=record["max_oob_percentage"],
oob_distance=record["oob_distance"])
states.append(sim_data_record)
simulation_data.params = SimulationParams(
beamng_steps=data["params"]["beamng_steps"],
delay_msec=int(data["params"]["delay_msec"]))
simulation_data.road = DecalRoad.from_dict(data["road"])
simulation_data.info = SimulationInfo()
simulation_data.info.start_time = data["info"]["start_time"]
simulation_data.info.end_time = data["info"]["end_time"]
simulation_data.info.elapsed_time = data["info"]["elapsed_time"]
simulation_data.info.success = data["info"]["success"]
simulation_data.info.computer_name = data["info"]["computer_name"]
simulation_data.info.id = data["info"]["id"]
simulation_data.states = states
if len(states) > 0:
member.distance_to_boundary = simulation_data.min_oob_distance()
member.simulation = simulation_data
individual: BeamNGIndividual = BeamNGIndividual(member, config)
#individual.evaluate()
b = tuple()
feature_dimensions = generate_feature_dimension(type)
for ft in feature_dimensions:
i = feature_simulator(ft.feature_simulator, individual)
b = b + (i, )
individuals.append([b, road, individual])
starting_point = choice(individuals)
original_set.append(starting_point)
i = 0
best_ind = individuals[0]
while i < popsize - 1:
max_dist = 0
for ind in individuals:
dist = get_min_distance_from_set(ind, original_set)
if dist > max_dist:
max_dist = dist
best_ind = ind
original_set.append(best_ind)
i += 1
base = config.initial_population_folder
storage = SeedStorage(base)
for index, road in enumerate(original_set):
dst = storage.get_path_by_index(index + 1)
ind = road[2]
#copy(road[1], dst)
with open(road[1]) as ff:
json_file = json.load(ff)
with open(dst, 'w') as f:
f.write(
json.dumps({
"control_nodes":
json_file["control_nodes"],
ind.m.simulation.f_params:
ind.m.simulation.params._asdict(),
ind.m.simulation.f_info:
ind.m.simulation.info.__dict__,
ind.m.simulation.f_road:
ind.m.simulation.road.to_dict(),
ind.m.simulation.f_records:
[r._asdict() for r in ind.m.simulation.states]
}))
# one step is 0.05 seconds (5/100)
beamng.step(STEPS)
try:
start()
finally:
beamng.close()
if __name__ == '__main__':
NODES = 20
MAX_ANGLE = 80
# MAX_ANGLE = 130
# MAX_ANGLE = 60
NUM_SPLINE_NODES = 20
SEG_LENGTH = 25
road = RoadGenerator(num_control_nodes=NODES, max_angle=MAX_ANGLE, seg_length=SEG_LENGTH,
num_spline_nodes=NUM_SPLINE_NODES).generate(visualise=False)
from self_driving.beamng_road_visualizer import plot_road
plot_road(road, save=True)
street = DecalRoad('street_1', drivability=1, material='').add_4d_points(road.sample_nodes)
run_sim(street)
def generate_random_solution_without_sim(self):
"""
To ease the bootstrap of the algorithm, we can generate
the first solutions in the feature space, so that we start
filling the bins
"""
# "Generate random solution"
path = self.problem._seed_pool_strategy.get_seed()
with open(path) as json_file:
data = json.load(json_file)
sample_nodes = data["road"]["nodes"]
for node in sample_nodes:
node[2] = -28.0
sample_nodes = np.array(sample_nodes)
records = data["records"]
bbox_size = (-250.0, 0.0, 250.0, 500.0)
road_bbox = RoadBoundingBox(bbox_size)
member = BeamNGMember(data["control_nodes"],
[tuple(t) for t in sample_nodes],
len(data["control_nodes"]), road_bbox)
member.config = self.config
member.problem = self.problem
simulation_id = time.strftime('%Y-%m-%d--%H-%M-%S', time.localtime())
sim_name = member.config.simulation_name.replace(
'$(id)', simulation_id)
simulation_data = SimulationData(sim_name)
states = []
for record in records:
state = VehicleState(timer=record["timer"],
pos=record["pos"],
dir=record["dir"],
vel=record["vel"],
steering=record["steering"],
steering_input=record["steering_input"],
brake=record["brake"],
brake_input=record["brake_input"],
throttle=record["throttle"],
throttle_input=record["throttle_input"],
wheelspeed=record["wheelspeed"],
vel_kmh=record["vel_kmh"])
sim_data_record = SimulationDataRecord(
**state._asdict(),
is_oob=record["is_oob"],
oob_counter=record["oob_counter"],
max_oob_percentage=record["max_oob_percentage"],
oob_distance=record["oob_distance"])
states.append(sim_data_record)
simulation_data.params = SimulationParams(
beamng_steps=data["params"]["beamng_steps"],
delay_msec=int(data["params"]["delay_msec"]))
simulation_data.road = DecalRoad.from_dict(data["road"])
simulation_data.info = SimulationInfo()
simulation_data.info.start_time = data["info"]["start_time"]
simulation_data.info.end_time = data["info"]["end_time"]
simulation_data.info.elapsed_time = data["info"]["elapsed_time"]
simulation_data.info.success = data["info"]["success"]
simulation_data.info.computer_name = data["info"]["computer_name"]
simulation_data.info.id = data["info"]["id"]
simulation_data.states = states
if len(states) > 0:
member.distance_to_boundary = simulation_data.min_oob_distance()
member.simulation = simulation_data
individual: BeamNGIndividual = BeamNGIndividual(
member, self.config)
individual.seed = path
else:
print("*********Bug************")
return individual
def run_sim(street_1: DecalRoad):
waypoints = []
for node in street_1.nodes:
waypoint = BeamNGWaypoint("waypoint_" + str(node),
get_node_coords(node))
waypoints.append(waypoint)
print(len(waypoints))
maps.beamng_map.generated().write_items(
street_1.to_json() + '\n' +
"\n".join([waypoint.to_json() for waypoint in waypoints]))
beamng = BeamNGpy('localhost', 64256)
scenario = Scenario('tig', 'tigscenario')
vehicle = Vehicle('ego_vehicle',
model='etk800',
licence='TIG',
color='Red')
sim_data_collector = TrainingDataCollectorAndWriter(
vehicle, beamng, street_1)
scenario.add_vehicle(vehicle,
pos=get_node_coords(street_1.nodes[0]),
rot=get_rotation(street_1))
scenario.make(beamng)
beamng.open()
beamng.set_deterministic()
beamng.load_scenario(scenario)
beamng.pause()
beamng.start_scenario()
vehicle.ai_drive_in_lane(True)
vehicle.ai_set_mode("disabled")
vehicle.ai_set_speed(10 / 4)
steps = 5
def start():
for waypoint in waypoints[10:-1:20]:
vehicle.ai_set_waypoint(waypoint.name)
for idx in range(1000):
if (idx * 0.05 * steps) > 3.:
sim_data_collector.collect_and_write_current_data()
dist = distance(sim_data_collector.last_state.pos,
waypoint.position)
if dist < 15.0:
beamng.resume()
break
# one step is 0.05 seconds (5/100)
beamng.step(steps)
try:
start()
finally:
beamng.close()
