def __init__(self, logdata=True, net=None):
self.steering_ctrl = CompositeController(
ProportionalController(0.4),
IntegrationController(0.2, integral_limit=1.5),
DerivativeController(2)
)
self.acceleration_ctrl = CompositeController(
ProportionalController(3.7),
)
self.data_logger = DataLogWriter() if logdata else None
self.counter = 0
# import the neural net that drives the car
self.model = simpleNetV2()
self.weights = 'simpleNetV2_epoch_3_all_tracks.csv.pkl'
self.model.load_state_dict(torch.load(self.weights))
self.input = torch.FloatTensor(D_in)
self.track_check1 = False
self.track_check2 = False
# import the neat neural network to handle the gear
config = neat.Config(neat.DefaultGenome, neat.DefaultReproduction,
neat.DefaultSpeciesSet, neat.DefaultStagnation,
'config-neat')
with open('winner-feedforward', 'rb') as f:
winner = pickle.load(f)
self.net = neat.nn.FeedForwardNetwork.create(winner, config)
self.clock = time.time()
self.done = False
# initialize the fitness with zero
self.temp_fitness = 0
def __init__(self, logdata=True):
# filenames = ['aalborg.csv', 'alpine-1.csv', 'f-speedway.csv', 'data_track_2.csv']
#filenames = ['aalborg.csv', 'alpine-1.csv', 'f-speedway.csv']
filenames = ['forza_urja.csv', 'newdata5.csv', 'aalborg_urja.csv']
#, 'aalborg_urja.csv', 'aalborg_urja_vx80.csv']
#filenames = ['aalborg_new.csv', 'forza_new.csv']
data, labels = helper_functions.readData(filenames)
# self.network = do_the_thing(data, labels)
learning_rate = 1e-6
#self.network = Network(data, labels, learning_rate)
#self.network.train()
#torch.save(self.network, 'current_network.pt')
self.network = torch.load('current_network_17_WINEF.pt')
self.id = random.uniform(0, 1)
fh = open("cooperation" + str(self.id) + ".txt", "w")
write(str(self.id) + ": 0.0")
fh.close()
self.set = False
self.steering_ctrl = CompositeController(
ProportionalController(0.4),
IntegrationController(0.2, integral_limit=1.5),
DerivativeController(2))
self.acceleration_ctrl = CompositeController(
ProportionalController(3.7), )
self.data_logger = DataLogWriter() if logdata else None
def __init__(self, logdata=True):
self.steering_ctrl = CompositeController(
ProportionalController(0.4),
IntegrationController(0.2, integral_limit=1.5),
DerivativeController(2))
self.acceleration_ctrl = CompositeController(
ProportionalController(3.7), )
self.data_logger = DataLogWriter() if logdata else None
# Setup the PyTorch model and load the weights
self.model = simpleNetV3()
weights = 'simpleNetV3_lr=0.01_epoch_1_all_tracks.csv.pkl'
self.model.load_state_dict(torch.load(weights))
# Initialize inputs, counters and history
self.input = torch.zeros(D_in)
self.crashCounter = 0
self.reverseCounter = 0
self.forwardCounter = 0
self.resetGear = False
self.crashed = False
self.counter = 0
self.name = '3001'
self.history = np.zeros((5, 2), dtype=float)
# Initialize SWARM
self.pheromones = []
pickle.dump(self.pheromones, open("sent_3001.txt", "wb"))
self.straight_begin = 0
self.straight_end = 0
self.list_straight = []
self.list_corner = []
self.received = []
self.list_average_corner = []
def __init__(self, logdata=True):
self.steering_ctrl = CompositeController(
ProportionalController(0.4),
IntegrationController(0.2, integral_limit=1.5),
DerivativeController(2))
self.acceleration_ctrl = CompositeController(
ProportionalController(3.7), )
self.data_logger = DataLogWriter() if logdata else None
def __init__(self, logdata=True, models=None, explore=0.9, optimizer=None):
self.steering_ctrl = CompositeController(
ProportionalController(0.4),
IntegrationController(0.2, integral_limit=1.5),
DerivativeController(2))
self.acceleration_ctrl = CompositeController(
ProportionalController(3.7), )
self.data_logger = DataLogWriter() if logdata else None
# Algorithm variables
self.previous_state = None
if os.path.isfile("memory.txt") == True:
with open("memory.txt", "rb") as fp:
self.replay_memory = pickle.load(fp)
else:
self.replay_memory = []
# Maximum size of the replay memory
self.max_RM_size = 100000
self.model, self.target = models
self.criterion = torch.nn.MSELoss()
self.model_optim = optimizer
self.restart = False
# Discretize the action space
self.action_space = [
0.9, 0.7, 0.5, 0.4, 0.3, 0.2, 0.1, 0, -0.1, -0.2, -0.3, -0.4, -0.5,
-0.7, -0.9
]
self.action = 0
self.reward = 0
# Model that is used for accelerating and braking
self.model_acc_brake = simpleNetV2()
self.pheromones = []
# self.receiver = Receiver(6001)
weights = 'simpleNetV2_epoch_3_all_tracks.csv.pkl'
self.model_acc_brake.load_state_dict(torch.load(weights))
self.input = torch.zeros(D_in)
# STATS
if os.path.isfile("counter.txt") == True:
with open("counter.txt", "rb") as fp:
self.counter = pickle.load(fp)
else:
self.counter = 0
self.counter_per_game = 0
self.train_counter = 0
self.average_reward = 0
self.average_loss = 0
# Hyperparameters
self.exploration_rate = explore
self.batch_size = 128
self.gamma = 0.99
def __init__(self, logdata=True):
self.steering_ctrl = CompositeController(
ProportionalController(0.4),
IntegrationController(0.2, integral_limit=1.5),
DerivativeController(2))
self.acceleration_ctrl = CompositeController(
ProportionalController(3.7), )
self.data_logger = DataLogWriter() if logdata else None
self.path = []
self.pid_steer = PID(Kp=0.5, Ki=0.0, Kd=1.0, setpoint=0, sample_time=0.02,\
output_limits=(-1.0, 1.0), auto_mode=True, proportional_on_measurement=True)
def __init__(self, logdata=True):
self.steering_ctrl = CompositeController(
ProportionalController(0.4),
IntegrationController(0.2, integral_limit=1.5),
DerivativeController(2)
)
self.acceleration_ctrl = CompositeController(
ProportionalController(3.7),
)
self.data_logger = DataLogWriter() if logdata else None
self.model = TwoLayerNet(D_in, hidden_size, hidden_size_2, D_out)
self.model.load_state_dict(torch.load('LaurensNet30.pkl'))
def instantiatePIDControllers(self):
self.steering_ctrl = CompositeController(
ProportionalController(-0.5),
IntegrationController(-0.1, integral_limit=1),
DerivativeController(-1))
self.reverse_ctrl = CompositeController(
ProportionalController(3.7),
# IntegrationController(0.1, integral_limit=1.5),
# DerivativeController(0.5)
)
self.forward_ctrl = CompositeController(
ProportionalController(3.7),
# IntegrationController(0.1, integral_limit=1.5),
# DerivativeController(0.5)
)
def __init__(self, logdata=True):
# filenames = ['aalborg.csv', 'alpine-1.csv', 'f-speedway.csv', 'data_track_2.csv']
#filenames = ['aalborg.csv', 'alpine-1.csv', 'f-speedway.csv']
filenames = ['forza_urja.csv']
#filenames = ['aalborg_new.csv', 'forza_new.csv']
data, labels = helper_functions.readData(filenames)
# self.network = do_the_thing(data, labels)
learning_rate = 1e-6
self.network = Network(data, labels, learning_rate)
self.network.train()
torch.save(self.network, 'current_network.pt')
self.steering_ctrl = CompositeController(
ProportionalController(0.4),
IntegrationController(0.2, integral_limit=1.5),
DerivativeController(2))
self.acceleration_ctrl = CompositeController(
ProportionalController(3.7), )
self.data_logger = DataLogWriter() if logdata else None
def __init__(self, logdata=True):
self.steering_ctrl = CompositeController(
ProportionalController(0.4),
IntegrationController(0.2, integral_limit=1.5),
DerivativeController(2))
self.acceleration_ctrl = CompositeController(
ProportionalController(3.7), )
self.data_logger = DataLogWriter() if logdata else None
self.model = simpleNetV2()
weights = 'simpleNetV2_epoch_3_all_tracks.csv.pkl'
self.model.load_state_dict(torch.load(weights))
self.input = torch.zeros(D_in)
self.crashCounter = 0
self.reverseCounter = 0
self.forwardCounter = 0
self.resetGear = False
self.crashed = False
self.counter = 0
self.name = '3001'
# NEAT
self.history = np.zeros((5, 2), dtype=float)
config = neat.Config(neat.DefaultGenome, neat.DefaultReproduction,
neat.DefaultSpeciesSet, neat.DefaultStagnation,
'config-neat')
with open('winner-feedforward', 'rb') as f:
winner = pickle.load(f)
self.net = neat.nn.FeedForwardNetwork.create(winner, config)
#SWARM
self.pheromones = []
pickle.dump(self.pheromones, open("../sent_3001.txt", "wb"))
self.straight_begin = 0
self.straight_end = 0
self.list_straight = []
self.list_corner = []
self.received = []
self.list_average_corcer = []
def __init__(self, logdata=True, net=None):
self.steering_ctrl = CompositeController(
ProportionalController(0.4),
IntegrationController(0.2, integral_limit=1.5),
DerivativeController(2)
)
self.acceleration_ctrl = CompositeController(
ProportionalController(3.7),
)
self.data_logger = DataLogWriter() if logdata else None
self.eta = 20
self.counter = 0
self.model = simpleNetV2()
self.weights = 'simpleNetV2_epoch_3_all_tracks.csv.pkl'
self.model.load_state_dict(torch.load(self.weights))
self.input = torch.FloatTensor(D_in)
self.track_check1 = False
self.track_check2 = False
self.action_neat = None
self.net = net
self.clock = time.time()
self.done = False
self.temp_fitness = 0
def __init__(self, logdata=False):
"""
set the initialization mode
"""
self.launch_torcs = True
self.launch_torcs_and_second_driver = False
"""
use_mlp_opponents: if True, the output of the ESN is adjusted by the Mulit-Layer Perceptron saved at PATH_TO_MLP
"""
self.use_mlp_opponents = True
self.PATH_TO_ESN = "./trained_nn/evesn10808.pkl" # for MLP evolution
#self.PATH_TO_ESN = "./trained_nn/simple_esn.pkl" # for reservoir evolution
self.PATH_TO_MLP = "./trained_nn/mlp_opponents.pkl"
self.CURRENT_SPEED_LIMIT = self.SPEED_LIMIT_NORMAL
"""
Controllers needed for the simple driver
"""
self.steering_ctrl = CompositeController(
ProportionalController(0.4),
IntegrationController(0.2, integral_limit=1.5),
DerivativeController(2))
self.acceleration_ctrl = CompositeController(
ProportionalController(3.7), )
self.data_logger = DataLogWriter() if logdata else None
"""
launch torcs practice race when ./start.sh is executed (if torcs is not already running)
comment out for normal use of torcs
"""
# automatically start TORCS when './start.sh' is executed
if self.launch_torcs:
torcs_command = [
"torcs", "-r",
os.path.abspath(
"./config_files/current_config_file/config.xml")
]
self.torcs_process = subprocess.Popen(torcs_command)
# automatically start TORCS with two drivers when './start.sh' is executed
elif self.launch_torcs_and_second_driver:
# assure that TORCS is not launched again when the second car is initialized
if not os.path.isfile('torcs_process.txt'):
torcs_command = [
"torcs", "-r",
os.path.abspath(
"./config_files/current_config_file/config.xml")
]
self.torcs_process = subprocess.Popen(torcs_command)
with open('./torcs_process.txt', 'w') as file:
file.write('running torcs process')
second_car_command = ["./start.sh", "-p", "3002"]
subprocess.Popen(second_car_command)
else:
line = ''
with open('./torcs_process.txt', 'r') as file:
line = file.readline()
if not line.find('running torcs process') > -1:
torcs_command = [
"torcs", "-r",
os.path.abspath(
"./config_files/current_config_file/config.xml")
]
self.torcs_process = subprocess.Popen(torcs_command)
with open('./torcs_process.txt', 'w') as file:
file.write('running torcs process')
second_car_command = ["./start.sh", "-p", "3002"]
subprocess.Popen(second_car_command)
else:
with open('./torcs_process.txt', 'w') as file:
file.write('')
"""
some global variables
"""
self.previous_position = 0
self.ID = 0
self.is_first = True
self.active_mlp = False
self.current_damage = 0
self.accel_deviation = 0
self.steering_deviation = 0