def test_single_agent_ring(self):
# create the base environment
env = FlowEnv(
flow_params=ring(
num_automated=5,
simulator="traci",
multiagent=False
),
multiagent=False,
shared=False,
version=1
)
env.reset()
# test observation space
test_space(
env.observation_space,
expected_min=np.array([-float("inf") for _ in range(25)]),
expected_max=np.array([float("inf") for _ in range(25)]),
expected_size=25,
)
# test action space
test_space(
env.action_space,
expected_min=np.array([-1 for _ in range(5)]),
expected_max=np.array([1 for _ in range(5)]),
expected_size=5,
)
# kill the environment
env.wrapped_env.terminate()
def setUp(self):
self.sim_params = deepcopy(ring())["sim"]
self.sim_params.render = False
# for AVClosedEnv
flow_params_closed = deepcopy(ring())
self.network_closed = flow_params_closed["network"](
name="test_closed",
vehicles=flow_params_closed["veh"],
net_params=flow_params_closed["net"],
)
self.env_params_closed = flow_params_closed["env"]
self.env_params_closed.additional_params = SA_CLOSED_ENV_PARAMS.copy()
# for AVOpenEnv
pass # TODO
def _get_ring_env_attributes(scale):
"""Return the environment parameters of the fast ring environment.
Parameters
----------
scale : int
the scale of the ring environment. The length of the network and the
number of human/RL vehicles is scaled by this value.
Returns
-------
dict
see ENV_ATTRIBUTES
"""
return {
"meta_ac_space":
lambda relative_goals, multiagent: Box(
low=-5 if relative_goals else 0,
high=5 if relative_goals else 10,
shape=(1 if multiagent else scale, ),
dtype=np.float32),
"state_indices":
lambda multiagent: [0]
if multiagent else [15 * i for i in range(scale)],
"env":
lambda evaluate, render, multiagent, shared, maddpg: FlowEnv(
flow_params=ring(
stopping_penalty=False,
acceleration_penalty=False,
scale=scale,
evaluate=evaluate,
multiagent=multiagent,
),
render=render,
multiagent=multiagent,
shared=shared,
maddpg=maddpg,
) if evaluate else
(RingMultiAgentEnv if multiagent else RingSingleAgentEnv)(
maddpg=maddpg,
length=[250 * scale, 360 * scale],
num_vehicles=22 * scale,
dt=0.2,
horizon=3000,
gen_emission=False,
rl_ids=[22 * i for i in range(scale)],
warmup_steps=0,
initial_state=os.path.join(
hbaselines_config.PROJECT_PATH,
"hbaselines/envs/mixed_autonomy/envs/initial_states/"
"ring-v{}.json".format(scale - 1)),
sims_per_step=1,
),
}
def test_multi_agent_ring(self):
# create the base environment
env = FlowEnv(
flow_params=ring(
num_automated=5,
simulator="traci",
multiagent=True
),
multiagent=True,
shared=False,
version=1
)
env.reset()
# test the agent IDs.
self.assertListEqual(
sorted(env.agents), ['rl_0_0', 'rl_0_1', 'rl_0_2', 'rl_0_3',
'rl_0_4'])
# test observation space
test_space(
env.observation_space["rl_0_0"],
expected_min=np.array([-float("inf") for _ in range(5)]),
expected_max=np.array([float("inf") for _ in range(5)]),
expected_size=5,
)
# test action space
test_space(
env.action_space["rl_0_0"],
expected_min=np.array([-1]),
expected_max=np.array([1]),
expected_size=1,
)
# kill the environment
env.wrapped_env.terminate()
# Mixed autonomy traffic flow environments. #
# ======================================================================= #
"ring-v0": {
"meta_ac_space":
lambda relative_goals, multiagent: Box(low=-5 if relative_goals else 0,
high=5
if relative_goals else 10,
shape=(5, ),
dtype=np.float32),
"state_indices":
lambda multiagent: [0],
"env":
lambda evaluate, render, n_levels, multiagent, shared, maddpg: FlowEnv(
flow_params=ring(
stopping_penalty=True,
acceleration_penalty=True,
evaluate=evaluate,
multiagent=multiagent,
),
render=render,
multiagent=multiagent,
shared=shared,
maddpg=maddpg,
),
},
"merge-v0": {
"meta_ac_space":
lambda relative_goals, multiagent: Box(low=-.5
if relative_goals else 0,
high=.5
if relative_goals else 1,
shape=(1
},
# ======================================================================= #
# Mixed autonomy traffic flow environments. #
# ======================================================================= #
"ring": {
"meta_ac_space":
lambda relative_goals: Box(low=-10 if relative_goals else 0,
high=10 if relative_goals else 30,
shape=(5, ),
dtype=np.float32),
"state_indices": [5 * i for i in range(5)],
"env":
lambda evaluate, render, multiagent, shared, maddpg: FlowEnv(
flow_params=ring(
evaluate=evaluate,
multiagent=multiagent,
),
render=render,
multiagent=multiagent,
shared=shared,
maddpg=maddpg,
),
},
"ring_small": {
"meta_ac_space":
lambda relative_goals: Box(low=-.5 if relative_goals else 0,
high=.5 if relative_goals else 1,
shape=(1, ),
dtype=np.float32),
"state_indices": [0],
"env":
# ======================================================================= #
# Mixed autonomy traffic flow environments. #
# ======================================================================= #
"ring": {
"meta_ac_space": lambda relative_goals: Box(
low=-1 if relative_goals else 0,
high=1,
shape=(5,),
dtype=np.float32
),
"state_indices": [5 * i for i in range(5)],
"env": lambda evaluate, render, multiagent, shared, maddpg: FlowEnv(
flow_params=ring(
evaluate=evaluate,
multiagent=multiagent,
),
render=render,
multiagent=multiagent,
shared=shared,
maddpg=maddpg,
),
},
"ring_small": {
"meta_ac_space": lambda relative_goals: Box(
low=-.5 if relative_goals else 0,
high=.5 if relative_goals else 1,
shape=(1,),
dtype=np.float32
),