def main():
conf = parser.parse_args()
in_path = conf.file_name
out_path = os.path.join(os.path.dirname(in_path), conf.output)
with open(in_path, 'r') as yamlfile:
yaml_bot = yamlfile.read()
body_spec = get_body_spec(conf)
brain_spec = get_extended_brain_spec(conf)
print "converting to protobuf..."
pb_bot = yaml_to_robot(body_spec, brain_spec, yaml_bot)
cpg_factory = CPG_Factory(body_spec=body_spec, brain_spec=brain_spec)
loopback = conf.loopback
print "Loopback: {0}".format(loopback)
coupling = conf.coupling
print "Coupling: {0}".format(coupling)
cpg_factory.add_CPGs(pb_bot, loopback, coupling)
print "converting to yaml..."
yaml_bot = robot_to_yaml(body_spec, brain_spec, pb_bot)
with open(out_path, 'w') as out_file:
out_file.write(yaml_bot)
print "done"
def main():
conf = parser.parse_args()
in_path = conf.file_name
out_path = os.path.join(os.path.dirname(in_path), conf.output)
with open(in_path, 'r') as yamlfile:
yaml_bot = yamlfile.read()
body_spec = get_body_spec(conf)
brain_spec = get_brain_spec(conf)
print "converting to protobuf..."
pb_bot = yaml_to_robot(body_spec, brain_spec, yaml_bot)
cpg_factory = CPG_Factory(body_spec=body_spec, brain_spec=brain_spec)
cpg_factory.add_CPGs(pb_bot, conf.type)
print "converting to yaml..."
yaml_bot = robot_to_yaml(body_spec, brain_spec, pb_bot)
with open(out_path, 'w') as out_file:
out_file.write(yaml_bot)
print "done"
def run(self, conf):
conf.min_parts = 1
conf.max_parts = 3
conf.arena_size = (3, 3)
conf.max_lifetime = 99999
conf.initial_age_mu = 99999
conf.initial_age_sigma = 1
conf.age_cutoff = 99999
self.body_spec = get_body_spec(conf)
self.brain_spec = get_brain_spec(conf)
self.nn_parser = NeuralNetworkParser(self.brain_spec)
print "OPENING FILES!!!!!!!!!!!!!!!!!!!"
with open("body/{0}".format(self.body_file), 'r') as robot_file:
robot_yaml = robot_file.read()
for filename in self.brain_files:
with open("brains/{0}".format(filename, 'r')) as brain_file:
br_yaml = brain_file.read()
self.brain_genotypes.append(
yaml_to_genotype(br_yaml, self.brain_spec))
yield From(wait_for(self.pause(True)))
pose = Pose(position=Vector3(0, 0, 0.5), rotation=rotate_vertical(0))
robot_pb = yaml_to_robot(self.body_spec, self.brain_spec, robot_yaml)
tree = Tree.from_body_brain(robot_pb.body, robot_pb.brain,
self.body_spec)
print "INSERTING ROBOT!!!!!!!!!!!!!!!!!!!!!!"
robot = yield From(wait_for(self.insert_robot(tree, pose)))
self.robot_name = robot.name
self.modify_nn_publisher = yield From(
self.manager.advertise(
'/gazebo/default/{0}/modify_neural_network'.format(
self.robot_name),
'gazebo.msgs.ModifyNeuralNetwork',
))
# Wait for connections
yield From(self.modify_nn_publisher.wait_for_listener())
brain_num = 0
num_of_brains = len(self.brain_genotypes)
print "Number of brains = {0}".format(num_of_brains)
yield From(wait_for(self.pause(False)))
while (True):
if self.timers.is_it_time('evaluate', self.time_period,
self.get_world_time()):
n = brain_num % num_of_brains
print "Switching brain to #{0}!!!!!!!!!".format(n)
yield From(self.insert_brain(self.brain_genotypes[n]))
self.timers.reset('evaluate', self.get_world_time())
brain_num += 1
yield From(trollius.sleep(0.1))
import uuid
from revolve.spec.msgs import Robot
import os
import sys
sys.path.append(os.path.dirname(os.path.abspath(__file__))+'/../')
from tol.spec import get_body_spec
from tol.config import parser, make_revolve_config
from sdfbuilder.math import Vector3
conf = make_revolve_config(parser.parse_args())
body_spec = get_body_spec(conf)
print("exp,run,robot_id,gene_id,origin_id,type,r,g,b")
def find_part_uuid(type, vec, parent_parts):
for pid, ptype, pvec, porigin in parent_parts:
if ptype == type and abs(vec - pvec) < 1e-8:
return pid, porigin
return None, None
def process_parts(robot_id, part, parent_parts, cur=None):
# Last three parameters are always color
type, vec = part.type, Vector3(*[p.value for p in part.param[-3:]])
pid, porigin = find_part_uuid(type, vec, parent_parts)
if pid is None:
pid, porigin = uuid.uuid4(), robot_id
import sys
import os
import glob
sys.path.append(os.path.dirname(os.path.abspath(__file__)) + '/../')
from revolve.angle.representation import Tree
from revolve.spec import Robot
from tol.spec import get_body_spec
from tol.config import parser, make_revolve_config
from tol.util.analyze import list_extremities, count_joints, count_motors
conf = parser.parse_args([])
make_revolve_config(conf)
body_spec = get_body_spec(conf)
input_dirs = sys.argv[1:]
for input_dir in input_dirs:
input_dir = os.path.abspath(input_dir)
print("Processing %s..." % input_dir)
files = glob.glob(os.path.join(input_dir, "*.pb"))
with open(os.path.join(input_dir, "robot_details.csv"), 'w') as o:
o.write(
"robot_id,size,extremity_id,extremity_size,joint_count,motor_count\n"
)
for filename in files:
robot = Robot()
with open(filename, "rb") as f:
robot.ParseFromString(f.read())
async def run():
conf = parser.parse_args()
conf.evaluation_time_sigma = 2.0
conf.weight_mutation_probability = 0.8
conf.weight_mutation_sigma = 5.0
conf.param_mutation_probability = 0.8
conf.param_mutation_sigma = 0.25
# this is the world state update frequency in simulation Hz
conf.pose_update_frequency = 5 # in simulation Hz
# update frequency of sensors in simulation Hz (default 10Hz)
# conf.sensor_update_rate = 10.
# # these are irrelevant parameters but we need to set them anyway,
# # otherwise it won't work
# conf.min_parts = 1
# conf.max_parts = 3
# conf.arena_size = (3, 3)
# conf.max_lifetime = 99999
# conf.initial_age_mu = 99999
# conf.initial_age_sigma = 1
# conf.age_cutoff = 99999
# create the learning manager
world = await LearningManager.create(conf)
await world.pause(True)
path_to_log_dir = os.path.join(world.path_to_log_dir, "learner1")
body_spec = get_body_spec(conf)
brain_spec = get_extended_brain_spec(conf)
# mutation spec contains info about what parameters of neurons can be mutated
mut_spec = get_extended_mutation_spec(conf.param_mutation_sigma,
conf.weight_mutation_sigma)
# what types of neurons can be added to the network
# allowed_types = ["Simple", "Sigmoid", "DifferentialCPG"]
# allowed_types = ["Simple", "Sigmoid"]
allowed_types = ["Simple"]
mutator = Mutator(mut_spec, allowed_neuron_types=allowed_types)
# if we are not restoring a saved state:
if not world.do_restore:
with open(conf.test_bot, 'r') as yamlfile:
bot_yaml = yamlfile.read()
pose = Pose(position=Vector3(0, 0, 0.2))
bot = yaml_to_robot(body_spec, brain_spec, bot_yaml)
tree = Tree.from_body_brain(bot.body, bot.brain, body_spec)
robot = await world.insert_robot(tree, pose)
learner = RobotLearner(world=world,
robot=robot,
brain_spec=brain_spec,
mutator=mutator,
conf=conf)
gen_files = []
init_brain_list = None
if (os.path.isdir(path_to_log_dir)):
gen_files = list(fname for fname in os.listdir(path_to_log_dir) if \
fnmatch.fnmatch(fname, "gen_*_genotypes.log"))
# if we are reading an initial population from a file:
if len(gen_files) > 0:
gen_files = sorted(gen_files,
key=lambda item: int(item.split('_')[1]))
last_gen_file = gen_files[-1]
num_generations = int(last_gen_file.split('_')[1]) # + 1
num_brains_evaluated = conf.population_size * num_generations
print("last generation file = {0}".format(last_gen_file))
# get list of brains from the last generation log file:
init_brain_list, min_mark, max_mark = \
get_brains_from_file(os.path.join(path_to_log_dir, last_gen_file), brain_spec)
print("Max historical mark = {0}".format(max_mark))
# set mutator's innovation number according to the max historical mark:
mutator.innovation_number = max_mark + 1
learner.total_brains_evaluated = num_brains_evaluated
learner.generation_number = num_generations
# initialize learner with initial list of brains:
await world.add_learner(learner, "learner1", init_brain_list)
print(learner.parameter_string())
# log experiment parameter values:
create_param_log_file(conf, learner.generation_number,
os.path.join(path_to_log_dir, "parameters.log"))
# copy the robot body file:
shutil.copy(conf.test_bot, path_to_log_dir)
# if we are restoring a saved state:
else:
print("WORLD RESTORED FROM {0}".format(world.world_snapshot_filename))
print("STATE RESTORED FROM {0}".format(world.snapshot_filename))
print("WORLD CREATED")
await world.run()
