def test_grid_size():
env = Warehouse(
shelf_columns=1,
column_height=3,
shelf_rows=3,
n_agents=1,
msg_bits=0,
sensor_range=1,
request_queue_size=5,
max_inactivity_steps=None,
max_steps=None,
reward_type=RewardType.GLOBAL,
)
assert env.grid_size == (14, 4)
env = Warehouse(
shelf_columns=3,
column_height=3,
shelf_rows=3,
n_agents=1,
msg_bits=0,
sensor_range=1,
request_queue_size=5,
max_inactivity_steps=None,
max_steps=None,
reward_type=RewardType.GLOBAL,
)
assert env.grid_size == (14, 10)
def test_goal_4(env_0: Warehouse):
assert env_0.request_queue[0] == env_0.shelfs[0]
_, rewards, _, _ = env_0.step([Action.LEFT])
assert rewards[0] == pytest.approx(0.0)
_, rewards, _, _ = env_0.step([Action.LEFT])
assert rewards[0] == pytest.approx(0.0)
_, rewards, _, _ = env_0.step([Action.FORWARD])
assert env_0.agents[0].x == 4
assert env_0.agents[0].y == 26
assert env_0.request_queue[0] == env_0.shelfs[0]
assert rewards[0] == pytest.approx(0.0)
def test_obs_space_2():
env = Warehouse(
shelf_columns=1,
column_height=3,
shelf_rows=3,
n_agents=10,
msg_bits=5,
sensor_range=1,
request_queue_size=5,
max_inactivity_steps=None,
max_steps=None,
reward_type=RewardType.GLOBAL,
)
obs = env.reset()
for s, o in zip(env.observation_space, obs):
assert len(gym.spaces.flatten(s, o)) == env._obs_length
def test_obs_space_1():
env = Warehouse(
shelf_columns=1,
column_height=3,
shelf_rows=3,
n_agents=10,
msg_bits=5,
sensor_range=1,
request_queue_size=5,
max_inactivity_steps=None,
max_steps=None,
reward_type=RewardType.GLOBAL,
)
obs = env.reset()
for _ in range(200):
obs, _, _, _ = env.step(env.action_space.sample())
assert env.observation_space.contains(obs)
def test_goal_2(env_1: Warehouse):
assert env_1.request_queue[0] == env_1.shelfs[0]
_, rewards, _, _ = env_1.step([Action.FORWARD, Action.NOOP])
assert env_1.agents[0].x == 4
assert env_1.agents[0].y == 28
assert env_1.request_queue[0] != env_1.shelfs[0]
assert rewards[0] == pytest.approx(1.0)
assert rewards[1] == pytest.approx(1.0)
def test_time_limit(time_limit):
env = Warehouse(
shelf_columns=1,
column_height=3,
shelf_rows=3,
n_agents=10,
msg_bits=5,
sensor_range=1,
request_queue_size=5,
max_inactivity_steps=None,
max_steps=time_limit,
reward_type=RewardType.GLOBAL,
)
_ = env.reset()
for _ in range(time_limit - 1):
_, _, done, _ = env.step(env.action_space.sample())
assert done == 10 * [False]
_, _, done, _ = env.step(env.action_space.sample())
assert done == 10 * [True]
def env_0():
env = Warehouse(3, 8, 3, 1, 0, 1, 5, 10, None, RewardType.GLOBAL)
env.reset()
env.agents[0].x = 4 # should place it in the middle (empty space)
env.agents[0].y = 27
env.agents[0].dir = Direction.DOWN
env.shelfs[0].x = 4
env.shelfs[0].y = 27
env.agents[0].carrying_shelf = env.shelfs[0]
env.request_queue[0] = env.shelfs[0]
env._recalc_grid()
return env
def test_action_space_1():
env = Warehouse(
shelf_columns=1,
column_height=3,
shelf_rows=3,
n_agents=2,
msg_bits=1,
sensor_range=1,
request_queue_size=5,
max_inactivity_steps=None,
max_steps=None,
reward_type=RewardType.GLOBAL,
)
env.reset()
assert env.action_space == spaces.Tuple(
2 * (spaces.MultiDiscrete([len(Action), 2]), ))
env.step(env.action_space.sample())
def env_double_agent_with_msg():
env = Warehouse(3, 8, 3, 2, 2, 1, 5, None, None, RewardType.GLOBAL)
env.reset()
return env
def env_single_agent():
env = Warehouse(3, 8, 3, 1, 0, 1, 5, None, None, RewardType.GLOBAL)
env.reset()
return env
def test_fast_obs_2():
env = Warehouse(3,
8,
3,
3,
2,
1,
5,
10,
None,
RewardType.GLOBAL,
fast_obs=False)
env.reset()
slow_obs_space = env.observation_space
for _ in range(10):
slow_obs = [env._make_obs(agent) for agent in env.agents]
env._use_fast_obs()
fast_obs = [env._make_obs(agent) for agent in env.agents]
assert len(fast_obs) == 3
assert len(slow_obs) == 3
flattened_slow = [
spaces.flatten(osp, obs)
for osp, obs in zip(slow_obs_space, slow_obs)
]
for i in range(len(fast_obs)):
print(slow_obs[0])
assert list(fast_obs[i]) == list(flattened_slow[i])
env._use_slow_obs()
env.step(env.action_space.sample())
def env_five_agents():
env = Warehouse(3, 8, 3, 5, 0, 1, 5, None, None, RewardType.GLOBAL)
env.reset()
return env