def run_scenario(vehicle_model='etk800',
deflation_pattern=[0, 0, 0, 0],
parts_config='vehicles/hopper/custom.pc',
run_number=0):
global base_filename, default_color, default_scenario, default_spawnpoint, steps_per_sec
global integral, prev_error, setpoint
integral = 0.0
prev_error = 0.0
# setup DNN model + weights
sm = DAVE2Model()
# steering_model = Model().define_model_BeamNG("BeamNGmodel-racetracksteering8.h5")
# throttle_model = Model().define_model_BeamNG("BeamNGmodel-racetrackthrottle8.h5")
dual_model = sm.define_dual_model_BeamNG()
# dual_model = sm.load_weights("BeamNGmodel-racetrackdualcomparison10K.h5")
# dual_model = sm.define_multi_input_model_BeamNG()
# dual_model = sm.load_weights("BeamNGmodel-racetrackdual-comparison10K-PIDcontrolset-4.h5")
# dual_model = sm.load_weights("BeamNGmodel-racetrackdual-comparison40K-PIDcontrolset-1.h5")
# BeamNGmodel-racetrack-multiinput-dualoutput-comparison10K-PIDcontrolset-1.h5
# BeamNGmodel-racetrackdual-comparison100K-PIDcontrolset-2
dual_model = sm.load_weights(
"BeamNGmodel-racetrackdual-comparison100K-PIDcontrolset-2.h5")
# dual_model = sm.load_weights("BeamNGmodel-racetrack-multiinput-dualoutput-comparison103K-PIDcontrolset-1.h5")
random.seed(1703)
setup_logging()
beamng = BeamNGpy('localhost',
64256,
home='H:/BeamNG.research.v1.7.0.1clean',
user='******')
scenario = Scenario(default_scenario, 'research_test')
vehicle = Vehicle('ego_vehicle',
model=vehicle_model,
licence='EGO',
color=default_color)
vehicle = setup_sensors(vehicle)
spawn = spawn_point(default_scenario, default_spawnpoint)
scenario.add_vehicle(
vehicle, pos=spawn['pos'], rot=None,
rot_quat=spawn['rot_quat']) #, partConfig=parts_config)
add_barriers(scenario)
# Compile the scenario and place it in BeamNG's map folder
scenario.make(beamng)
# Start BeamNG and enter the main loop
bng = beamng.open(launch=True)
#bng.hide_hud()
bng.set_deterministic() # Set simulator to be deterministic
bng.set_steps_per_second(steps_per_sec) # With 60hz temporal resolution
bng.load_scenario(scenario)
bng.start_scenario()
# bng.spawn_vehicle(vehicle, pos=spawn['pos'], rot=None, rot_quat=spawn['rot_quat'], partConfig=parts_config)
# Put simulator in pause awaiting further inputs
bng.pause()
assert vehicle.skt
# bng.resume()
# perturb vehicle
print("vehicle position before deflation via beamstate:{}".format(
vehicle.get_object_position()))
print("vehicle position before deflation via vehicle state:{}".format(
vehicle.state))
image = bng.poll_sensors(vehicle)['front_cam']['colour'].convert('RGB')
# plt.imshow(image)
# plt.pause(0.01)
vehicle.deflate_tires(deflation_pattern)
bng.step(steps_per_sec * 6)
vehicle.update_vehicle()
# print("vehicle position after deflation via beamstate:{}".format(vehicle.get_object_position()))
# print("vehicle position after deflation via vehicle state:{}".format(vehicle.state))
pitch = vehicle.state['pitch'][0]
roll = vehicle.state['roll'][0]
z = vehicle.state['pos'][2]
image = bng.poll_sensors(vehicle)['front_cam']['colour'].convert('RGB')
# plt.imshow(image)
# plt.pause(0.01)
# bng.resume()
# vehicle.break_all_breakgroups()
# vehicle.break_hinges()
wheelspeed = 0.0
throttle = 0.0
prev_error = setpoint
damage_prev = None
runtime = 0.0
kphs = []
traj = []
pitches = []
rolls = []
steering_inputs = []
throttle_inputs = []
timestamps = []
damage = None
final_img = None
# Send random inputs to vehice and advance the simulation 20 steps
overall_damage = 0.0
total_loops = 0
total_imgs = 0
total_predictions = 0
while overall_damage <= 0:
# collect images
sensors = bng.poll_sensors(vehicle)
image = sensors['front_cam']['colour'].convert('RGB')
# plt.imshow(image)
# plt.pause(0.01)
total_imgs += 1
img = sm.process_image(np.asarray(image))
wheelspeed = sensors['electrics']['wheelspeed']
kph = ms_to_kph(wheelspeed)
dual_prediction = dual_model.predict(x=[img, np.array([kph])])
# steering_prediction = steering_model.predict(img)
# throttle_prediction = throttle_model.predict(img)
dt = sensors['timer']['time'] - runtime
runtime = sensors['timer']['time']
# control params
brake = 0
# steering = float(steering_prediction[0][0]) #random.uniform(-1.0, 1.0)
# throttle = float(throttle_prediction[0][0])
steering = float(dual_prediction[0][0]) #random.uniform(-1.0, 1.0)
throttle = float(dual_prediction[0][1])
total_predictions += 1
if abs(steering) > 0.2:
setpoint = 20
else:
setpoint = 40
# if runtime < 10:
throttle = throttle_PID(kph, dt)
# if throttle > 1:
# throttle = 1
# if setpoint < kph:
# brake = 0.0 #throttle / 10000.0
# throttle = 0.0
vehicle.control(throttle=throttle, steering=steering, brake=brake)
steering_inputs.append(steering)
throttle_inputs.append(throttle)
timestamps.append(runtime)
steering_state = sensors['electrics']['steering']
steering_input = sensors['electrics']['steering_input']
avg_wheel_av = sensors['electrics']['avg_wheel_av']
damage = sensors['damage']
overall_damage = damage["damage"]
new_damage = diff_damage(damage, damage_prev)
damage_prev = damage
vehicle.update_vehicle()
traj.append(vehicle.state['pos'])
pitches.append(vehicle.state['pitch'][0])
rolls.append(vehicle.state['roll'][0])
kphs.append(ms_to_kph(wheelspeed))
total_loops += 1
if new_damage > 0.0:
final_img = image
break
bng.step(1, wait=True)
if runtime > 300:
print("Exited after 5 minutes successful runtime")
break
if distance(
spawn['pos'],
vehicle.state['pos']) < 5 and sensors['timer']['time'] > 10:
reached_start = True
break
# print("runtime:{}".format(round(runtime, 2)))
# print("time to crash:{}".format(round(runtime, 2)))
bng.close()
# avg_kph = float(sum(kphs)) / len(kphs)
plt.imshow(final_img)
plt.savefig("Run-{}-finalimg.png".format(run_number))
plt.pause(0.01)
plot_input(timestamps, steering_inputs, "Steering", run_number=run_number)
plot_input(timestamps, throttle_inputs, "Throttle", run_number=run_number)
plot_input(timestamps, kphs, "KPHs", run_number=run_number)
print("Number of steering_inputs:", len(steering_inputs))
print("Number of throttle inputs:", len(throttle_inputs))
results = "Total loops: {} \ntotal images: {} \ntotal predictions: {} \nexpected predictions ={}*{}={}".format(
total_loops, total_imgs, total_predictions, round(runtime, 3),
steps_per_sec, round(runtime * steps_per_sec, 3))
print(results)
results = {
'runtime': round(runtime, 3),
'damage': damage,
'kphs': kphs,
'traj': traj,
'pitch': round(pitch, 3),
'roll': round(roll, 3),
"z": round(z, 3),
'final_img': final_img,
'total_predictions': total_predictions,
'expected_predictions': round(runtime * steps_per_sec, 3)
}
return results
def run_scenario_ai_version(vehicle_model='etk800',
deflation_pattern=[0, 0, 0, 0],
parts_config=None):
global base_filename, default_color, default_scenario, default_spawnpoint, setpoint, steps_per_sec
global prev_error
random.seed(1703)
setup_logging()
beamng = BeamNGpy('localhost',
64256,
home='H:/BeamNG.research.v1.7.0.1clean',
user="******")
scenario = Scenario(default_scenario, 'research_test')
vehicle = Vehicle('ego_vehicle',
model=vehicle_model,
licence='AI',
color=default_color)
vehicle = setup_sensors(vehicle)
spawn = get_spawn_point(default_scenario, default_spawnpoint)
# scenario.add_vehicle(vehicle, pos=spawn['pos'], rot=None, rot_quat=spawn['rot_quat'])
# Compile the scenario and place it in BeamNG's map folder
scenario.make(beamng)
# Start BeamNG and enter the main loop
bng = beamng.open(launch=True)
bng.set_deterministic() # Set simulator to be deterministic
bng.set_steps_per_second(steps_per_sec) # With 100hz temporal resolution
# Load and start the scenario
bng.load_scenario(scenario)
bng.start_scenario()
# create vehicle to be chased
chase_vehicle = Vehicle('chase_vehicle',
model='miramar',
licence='CHASEE',
color='Red')
bng.spawn_vehicle(chase_vehicle,
pos=(469.784, 346.391, 144.982),
rot=None,
rot_quat=(-0.0037852677050978, -0.0031219546217471,
-0.78478640317917, 0.61974692344666))
bng.spawn_vehicle(vehicle,
pos=spawn['pos'],
rot=None,
rot_quat=spawn['rot_quat'],
partConfig=parts_config)
# Put simulator in pause awaiting further inputs
bng.pause()
assert vehicle.skt
#bng.resume()
# perturb vehicle
print("vehicle position before deflation via beamstate:{}".format(
vehicle.get_object_position()))
print("vehicle position before deflation via vehicle state:{}".format(
vehicle.state))
image = bng.poll_sensors(vehicle)['front_cam']['colour'].convert('RGB')
plt.imshow(image)
plt.pause(0.01)
vehicle.deflate_tires(deflation_pattern)
bng.step(steps_per_sec * 6)
vehicle.update_vehicle()
# print("vehicle position after deflation via beamstate:{}".format(vehicle.get_object_position()))
# print("vehicle position after deflation via vehicle state:{}".format(vehicle.state))
pitch = vehicle.state['pitch'][0]
roll = vehicle.state['roll'][0]
z = vehicle.state['pos'][2]
image = bng.poll_sensors(vehicle)['front_cam']['colour'].convert('RGB')
plt.imshow(image)
plt.pause(0.01)
bng.resume()
vehicle.ai_set_mode('chase')
vehicle.ai_set_target('chase_vehicle')
vehicle.ai_drive_in_lane(True)
damage_prev = None
runtime = 0.0
traj = []
kphs = []
for _ in range(650):
image = bng.poll_sensors(vehicle)['front_cam']['colour'].convert('RGB')
damage = bng.poll_sensors(vehicle)['damage']
wheelspeed = bng.poll_sensors(vehicle)['electrics']['wheelspeed']
new_damage = diff_damage(damage, damage_prev)
damage_prev = damage
runtime = bng.poll_sensors(vehicle)['timer']['time']
vehicle.update_vehicle()
traj.append(vehicle.state['pos'])
kphs.append(ms_to_kph(wheelspeed))
if new_damage > 0.0:
break
bng.step(5)
bng.close()
results = {
'runtime': round(runtime, 3),
'damage': damage,
'kphs': kphs,
'traj': traj,
'pitch': round(pitch, 3),
'roll': round(roll, 3),
"z": round(z, 3),
'final_img': image
}
return results
