def test_wall(self):
env = fse.FoodSearch(obs_horizon=2, n_noise_channels=2)
self.assertListEqual(
env.board[env.wall_distance_from_boundary,
env.wall_distance_from_boundary:-env.wall_distance_from_boundary,
env.wall_channel].tolist(),
[1]*(env.board_size - 2*env.wall_distance_from_boundary)
)
self.assertListEqual(
env.board[-env.wall_distance_from_boundary - 1,
env.wall_distance_from_boundary:-env.wall_distance_from_boundary,
env.wall_channel].tolist(),
[1]*(env.board_size - 2*env.wall_distance_from_boundary)
)
self.assertListEqual(
env.board[env.wall_distance_from_boundary:-env.wall_distance_from_boundary,
env.wall_distance_from_boundary,
env.wall_channel].tolist(),
[1]*(env.board_size - 2*env.wall_distance_from_boundary)
)
self.assertListEqual(
env.board[env.wall_distance_from_boundary:-env.wall_distance_from_boundary,
-env.wall_distance_from_boundary - 1,
env.wall_channel].tolist(),
[1]*(env.board_size - 2*env.wall_distance_from_boundary)
)
def test_free_agent_neighborhood(self):
env = fse.FoodSearch(obs_horizon=1, n_noise_channels=2)
for i in range(1000):
observation = env.reset()
# No great apples in observation
self.assertTrue(
np.all(observation[:, :, 0] == 0),
msg=f'{observation[:, :, 0]}'
)
def test_rewards(self):
env = fse.FoodSearch(obs_horizon=2, n_noise_channels=2)
observation = env.reset()
env.board = np.zeros_like(env.board)
env.agent_position = np.array([8, 8])
env.board[7, 8, 0] = 1
observation, reward, done, _ = env.step(0) # North
self.assertTrue(done)
self.assertEqual(reward, env.apple_rewards[0])
def test_reset(self):
env = fse.FoodSearch(obs_horizon=2, n_noise_channels=2,
rng=np.random.RandomState(1234567))
observation = env.reset()
self.assertEqual(observation.shape, env.observation_space.shape)
self.assertTrue(
all(x in {0, 1} for x in observation.flatten())
)
# Object encoding part of board has to have unique object at each location
self.assertTrue(
np.all(np.sum(env.board[:, :, :len(env.apple_rewards)],
axis=-1).flatten() <= 1)
)
self.assertTrue(
np.any(np.sum(env.board[:, :, :],
axis=-1).flatten() > 1)
)
self.assertListEqual(
env.board[3, 3].tolist(),
[1, 0, 0, 0, 1]
)
def test_actions(self):
env = fse.FoodSearch(obs_horizon=2, n_noise_channels=2)
observation = env.reset()
env.board = np.zeros_like(env.board)
env.agent_position = np.array([8, 8])
observation, reward, done, _ = env.step(0)
self.assertListEqual(env.agent_position.tolist(),
[7, 8])
self.assertFalse(done)
observation, reward, done, _ = env.step(1)
self.assertListEqual(env.agent_position.tolist(),
[7, 9])
self.assertFalse(done)
observation, reward, done, _ = env.step(2)
self.assertListEqual(env.agent_position.tolist(),
[8, 9])
self.assertFalse(done)
observation, reward, done, _ = env.step(3)
self.assertListEqual(env.agent_position.tolist(),
[8, 8])
self.assertFalse(done)