def main():
"""
Demo how to run an agent
"""
if( len(sys.argv) > 1 ):
configName = str(sys.argv[1])
filename = configName
else:
print("Default config ")
configName = "./configs/new_traffic_loop_ppo.yaml"
dirname = os.path.dirname(__file__)
filename = os.path.join(dirname, configName)
print( "Config name " + configName )
logging.info("Starting example PPO agent")
logoutput = io.StringIO("episode output log")
parameters = getParameters(filename)
env = SumoGymAdapter(parameters['all'])
# here we initialize all agents (in that case one)
PPOAgents = []
env.reset()
for intersectionId in env.action_space.spaces.keys():
PPOAgents.append(PPOAgent(agentId=intersectionId, environment=env, parameters=parameters['all']))
complexAgent = BasicComplexAgent(PPOAgents)
experiment = Experiment(complexAgent, env, parameters['all']['max_steps'], parameters['all']['seedlist'], render=False)
experiment.addListener(JsonLogger(logoutput))
experiment.run()
def main():
"""
Demo how to run an agent
"""
# if( len(sys.argv) > 1 ):
# configName = str(sys.argv[1])
# filename = configName
# else:
# print("Default config ")
# configName = "./configs/new_traffic_loop_ppo.yaml"
# dirname = os.path.dirname(__file__)
# filename = os.path.join(dirname, configName)
# print( "Config name " + configName )
logging.info("Starting example random agent")
logoutput = io.StringIO("episode output log")
# parameters = getParameters(filename)
env = Warehouse()
# here we initialize all agents (in that case one)
randomAgents = []
env.reset()
for agent_id in env.action_space.spaces.keys():
randomAgents.append(WarehouseNaiveAgent(robot_id=agent_id, env=env))
complexAgent = BasicComplexAgent(randomAgents)
experiment = Experiment(complexAgent,
env,
maxSteps=100,
render=True,
renderDelay=0.5)
experiment.addListener(JsonLogger(logoutput))
experiment.run()
def main():
"""
Demo how to run an agent
"""
dirname = os.path.dirname(__file__)
if (len(sys.argv) > 1):
env_configName = str(sys.argv[1])
agent_configName = str(sys.argv[2])
else:
print("Default config ")
env_configName = "./configs/factory_floor_experiment.yaml"
env_filename = os.path.join(dirname, env_configName)
agent_configName = "./configs/agent_combined_config.yaml"
agent_filename = os.path.join(dirname, agent_configName)
env_parameters = getParameters(env_filename)
agent_parameters = getParameters(agent_filename)
# whao, you need to know exact contents of all files here..
recursive_update(
agent_parameters['subAgentList'][0]['parameters']['simulator'],
env_parameters['environment'])
print(env_parameters)
print(agent_parameters)
random.seed(env_parameters['seed'])
maxSteps = env_parameters['max_steps']
baseEnv = FactoryFloor(env_parameters['environment'])
packedActionSpace = PackedSpace(baseEnv.action_space,
{"robots": ["robot1", "robot2", "robot3"]})
env = ModifiedGymEnv(baseEnv, packedActionSpace)
logging.info("Starting example MCTS agent")
logoutput = io.StringIO("episode output log")
try:
logoutputpickle = open('./' + os.environ["SLURM_JOB_ID"] + '.pickle',
'wb')
except KeyError:
print("No SLURM_JOB_ID found")
logoutputpickle = io.BytesIO()
obs = env.reset()
complexAgent = createAgent(env.action_space, env.observation_space,
agent_parameters)
experiment = Experiment(complexAgent, env, maxSteps, render=True)
experiment.addListener(JsonLogger(logoutput))
experiment.addListener(PickleLogger(logoutputpickle))
stats, confidence_ints = experiment.run()
logoutputpickle.close()
print("json output:", logoutput.getvalue())
print("\n\nREWARD STATS: " + str(stats) + " \nCONFIDENCE INTERVALS " +
str(confidence_ints))
def testRun1step(self):
agent = Mock()
env = Mock()
# each episode succeeds in step 1
env.step = Mock(return_value=('observation', 3.0, True, {}))
firstAction = Mock()
exp = Experiment(agent, env, 100, None, False, 0)
result = np.mean(exp.run())
self.assertEqual(3.0, result)
def testListener(self):
agent = Mock()
env = Mock()
env.step = Mock(return_value=('observation', 3.0, True, {}))
firstAction = Mock()
exp = Experiment(agent, env, 30, None, False, 0)
listener = Mock()
exp.addListener(listener)
result = exp.run()
# check we got 30 callbacks
self.assertEquals(30, len(listener.notifyChange.mock_calls))
def testListener(self):
agent = Mock()
env = Mock()
env.step = Mock(return_value=('observation', 3.0, True, {}))
firstAction = Mock()
exp = Experiment(agent, env, 30, None, False, 0)
listener = Mock()
exp.addListener(listener)
result = exp.run()
# check we got 30 callbacks
# This fails because Jinke added an entry to the notification dictionary in Experiment.py
self.assertEquals(30, len(listener.notifyChange.mock_calls))
def main():
if (len(sys.argv) > 1):
configName = str(sys.argv[1])
filename = configName
else:
print("Default config ")
configName = "../configs/PPO.yaml"
dirname = os.path.dirname(__file__)
filename = os.path.join(dirname, configName)
logging.info("Starting example warehouse DQN agent")
logoutput = io.StringIO("episode output log")
parameters = getParameters(filename)
env = Warehouse()
# here we initialize all agents (in that case one)
agents = []
env.reset()
for agent_id, action_space in env.action_space.spaces.items():
agents.append(
WarehouseNaiveAgent(robot_id=agent_id,
actionspace=action_space,
observationspace=env.observation_space))
breakpoint()
agents[20] = PPOAgent(agentId=20,
actionspace=env.action_space[20],
observationspace=env.observation_space,
parameters=parameters['all'])
multi_agent = MultiAgent(agents)
experiment = Experiment(
multi_agent,
env,
maxSteps=1e6,
render=True,
seedlist=[random.randrange(1, 1e3, 1) for i in range(int(1e5))],
renderDelay=0.5)
experiment.addListener(JsonLogger(logoutput))
experiment.run()
def testRun3steps(self):
agent = Mock()
env = Mock()
env.step = Mock()
# each episode succeeds after 3 steps, then restart 10*
env.step.side_effect = [('observation1', 3.0, False, {}),
('observation2', 4.0, False, {}),
('observation3', 5.0, True, {})] * 10
firstAction = Mock()
exp = Experiment(agent, env, 30, None, False, 0)
result = np.mean(exp.run())
# each episode has reward 12
self.assertEqual(12.0, result)
def test_PPO_agent(self):
logging.info("Starting test_PPO_agent")
dirname = os.path.dirname(__file__)
filename = os.path.join(dirname, "configs/new_traffic_loop_ppo.yaml")
with open(filename, 'r') as stream:
try:
parameters = yaml.safe_load(stream)['parameters']
except yaml.YAMLError as exc:
logging.error(exc)
env = SumoGymAdapter(parameters)
env.reset()
PPOAgents = []
for intersectionId in env.action_space.spaces.keys():
PPOAgents.append(
PPOAgent(intersectionId, env.action_space,
env.observation_space, parameters))
complexAgent = BasicComplexAgent(PPOAgents, env.action_space,
env.observation_space)
experiment = Experiment(complexAgent, env, parameters['max_steps'])
experiment.run()
def main():
"""
MCTS Factory Floor experiment
"""
dirname = os.path.dirname(__file__)
parser = configargparse.ArgParser()
parser.add('-e',
'--env-config',
dest="env_filename",
default=os.path.join(
dirname, "./debug_configs/factory_floor_experiment.yaml"))
parser.add('-a',
'--agent-config',
dest="agent_filename",
default=os.path.join(dirname,
"./debug_configs/agent_config.yaml"))
parser.add('-d', '--data-dirname', dest="data_dirname", default="data")
args = parser.parse_args()
try:
data_outputdir = os.path.join(
dirname,
"./" + args.data_dirname + "/" + os.environ["SLURM_JOB_ID"])
os.makedirs(data_outputdir)
logoutputpickle = open('./' + data_outputdir + '/output.pickle', 'wb')
rewardsFile = open('./' + data_outputdir + '/rewards.yaml', 'w+')
except KeyError:
print("No SLURM_JOB_ID found")
logoutputpickle = io.BytesIO()
rewardsFile = io.StringIO()
env_parameters = getParameters(args.env_filename)
agent_parameters = getParameters(args.agent_filename)
print(env_parameters)
print(agent_parameters)
for subAgent in agent_parameters["subAgentList"]:
subAgent["parameters"]["simulator"] = env_parameters["environment"]
subAgent["parameters"]["simulator"][
"fullname"] = "aienvs.FactoryFloor.FactoryFloor.FactoryFloor"
random.seed(env_parameters['seed'])
maxSteps = env_parameters['max_steps']
env = FactoryFloor(env_parameters['environment'])
logging.info("Starting example MCTS agent")
logoutput = io.StringIO("episode output log")
obs = env.reset()
complexAgent = createAgent(env.action_space, env.observation_space,
agent_parameters)
experiment = Experiment(complexAgent, env, maxSteps, render=False)
experiment.addListener(JsonLogger(logoutput))
experiment.addListener(PickleLogger(logoutputpickle))
rewards = experiment.run()
# statistics, confidence_ints = stats.describe(rewards), stats.bayes_mvs(rewards)
logoutputpickle.close()
yaml.dump(rewards, rewardsFile)
rewardsFile.close()
print("json output:", logoutput.getvalue())
print("\n\nREWARD STATS: " + str(statistics) + " \nCONFIDENCE INTERVALS " +
str(confidence_ints))