def main(args):
a_args = (args.pip, args.pport, args.fsm_config,)
a_args += ("usermodule", True, None, {"filename": args.file, "append": args.append},
ConfigMgr(args.user_config, "pygame", eval_key=True),)
client = None
if args.client == 'webots':
client = WebotsClient()
if args.exp == "penaltykick":
run(PenaltyKickEnvironment, PenaltyKickAgent, EpisodicExperiment,
env_args=(args.feature_rep, args.support_leg, client,),
agent_args=a_args,
exp_kwargs={'nepisodes': args.nepisodes},
local=args.local)
elif args.exp == "bodymotion":
run(BodyMotionEnvironment, BodyMotionAgent, EpisodicExperiment,
env_args=(args.feature_rep, client,),
agent_args=a_args,
exp_kwargs={'nepisodes': args.nepisodes},
local=args.local)
def main(args):
if args.single_pole:
if args.friction:
# single pole with friction
run(CartPole, ModelBasedAgent, EpisodicExperiment,
env_args=('custom', [.5], [.1], 10),
env_kwargs={'random_start': False, 'cart_loc': [-2.4, 2.4], 'cart_vel': [-np.inf, np.inf],
'pole_angle': [-36, 36], 'pole_vel': [-np.inf, np.inf], 'mu_c': 0.0005, 'mu_p': 0.000002,
'discount_factor': 0.99},
agent_args=('learningmodule', False, None, None, None, 'cacla', 0, 40),
agent_kwargs={'alpha': 0.001, 'beta': 0.001},
exp_args=(1, 1000, 1000),
exp_kwargs={'filename': args.file},
local=True)
else:
run(CartPole, ModelBasedAgent, EpisodicExperiment,
env_args=('custom', [.5], [.1], 10),
env_kwargs={'random_start': False, 'cart_loc': [-2.4, 2.4], 'cart_vel': [-np.inf, np.inf],
'pole_angle': [-36, 36], 'pole_vel': [-np.inf, np.inf], 'mu_c': 0.0005, 'mu_p': 0.000002,
'discount_factor': 0.99, 'include_friction': False},
agent_args=('learningmodule', False, None, None, None, 'cacla', 0, 40),
agent_kwargs={'alpha': 0.001, 'beta': 0.001},
exp_args=(1, 500, 1000),
exp_kwargs={'filename': args.file},
local=True)
else:
# double pole with friction
run(CartPole, ModelBasedAgent, EpisodicExperiment,
env_args=('custom', [1., .1], [.1, .01], 10),
env_kwargs={'discrete_action': False, 'random_start': True, 'cart_loc': [-2.4, 2.4],
'cart_vel': [-np.inf, np.inf], 'pole_angle': [-36., 36.], 'pole_vel': [-np.inf, np.inf],
'mu_c': 0.0005, 'mu_p': 0.000002, 'discount_factor': 0.99},
agent_args=('learningmodule', False, None, None, None, 'cacla', 0, 40),
agent_kwargs={'alpha': 0.001, 'beta': 0.001},
exp_args=(1, 200000, 1000),
exp_kwargs={'filename': args.file},
local=True)
{"filename": args.infile, "append": args.append},
args.learner,
np.asarray(demo),
ValueIteration(model, explorer, args.gamma, args.ignore_unreachable),
)
a_kwargs = {"filename": args.savefile, "progress": args.progress}
client = None
if args.client == "webots":
client = WebotsClient()
run(
PenaltyKickEnvironment,
PenaltyKickAgent,
EpisodicExperiment,
env_args=(args.feature_rep, args.support_leg, client),
agent_args=a_args,
agent_kwargs=a_kwargs,
local=args.local,
)
if __name__ == "__main__":
ap = argparse.ArgumentParser(description="Experiment: Emulate the behavior via inverse reinforcement learning.")
ap.add_argument(
"--local", action="store_false", help="When true, the experiment is run through a local RLGlue implementation."
)
ap.add_argument("--progress", action="store_true", help="When true, a previously saved trial will be continued.")
ap.add_argument(
"--keep_history",
action="store_true",
sys.exit(sys.exc_info()[1])
except KeyError, e:
sys.exit("Key not found: {0}".format(e))
a_args = (args.pip, args.pport, args.fsm_config,)
a_args += ("followpolicymodule", True, None, {"filename": args.file, "append": args.append},
policies,)
client = None
if args.client == 'webots':
client = WebotsClient()
if args.exp == "penaltykick":
run(PenaltyKickEnvironment, PenaltyKickAgent, EpisodicExperiment,
env_args=(args.feature_rep, args.support_leg, client,),
agent_args=a_args,
exp_kwargs={'ntrials': args.ntrials, 'nepisodes': args.nepisodes},
local=args.local)
elif args.exp == "bodymotion":
run(BodyMotionEnvironment, BodyMotionAgent, EpisodicExperiment,
env_args=(args.feature_rep, client,),
agent_args=a_args,
exp_kwargs={'ntrials': args.ntrials, 'nepisodes': args.nepisodes},
local=args.local)
if __name__ == "__main__":
ap = argparse.ArgumentParser(description="Gather observation data for the given task from the NAO robot.")
ap.add_argument("--local", action="store_false",
help="When true, the experiment is run through a local RLGlue implementation.")
ap.add_argument("--append", action="store_true",
