def test_taking_actions():
"""Does the environment correctly change the state when told to take an action with and without stochasticity?"""
random.seed()
env = Gridworld(4, 4, 0.0)
# Deterministic tests
assert env.next_state(env.initial_state(), Action.up) == env.state_from_grid_position(GridPosition(0, 1))
assert env.next_state(env.initial_state(), Action.down) == env.state_from_grid_position(GridPosition(0, 0))
assert env.next_state(env.initial_state(), Action.left) == env.state_from_grid_position(GridPosition(0, 0))
assert env.next_state(env.initial_state(), Action.right) == env.state_from_grid_position(GridPosition(1, 0))
# Stochastic tests
env.failure_rate = 0.1
assert ratio_test(lambda state: state == env.state_from_grid_position(GridPosition(0, 0)), partial(env.next_state, env.initial_state(), Action.left), 10000) == 1.0
ratio = ratio_test(lambda state: state == env.state_from_grid_position(GridPosition(0, 0)), partial(env.next_state, env.initial_state(), Action.up), 10000)
assert 0.09 < ratio < 0.11
def test_taking_actions():
"""Does the environment correctly change the state when told to take an action with and without stochasticity?"""
random.seed()
env = GridworldContinuous(0.05, 0.01)
start = env.initial_state()
ratio = ratio_test(lambda state: np.linalg.norm(np.asarray([state[0] - start[0], state[1] - (start[1] + env.move_mean)]), 2) < env.move_sd * 2,
partial(env.next_state, start, Action.up), 10000)
assert 0.7 < ratio
steps = 0
s = env.initial_state()
while not env.is_terminal(s):
s = env.next_state(s, np.random.randint(4))
steps += 1
assert steps < 20000
def test_taking_actions():
"""Does the environment correctly change the state when told to take an action with and without stochasticity?"""
random.seed()
env = GridworldContinuous(0.05, 0.01)
start = env.initial_state()
ratio = ratio_test(
lambda state: np.linalg.norm(
np.asarray(
[state[0] - start[0], state[1] -
(start[1] + env.move_mean)]), 2) < env.move_sd * 2,
partial(env.next_state, start, Action.up), 10000)
assert 0.7 < ratio
steps = 0
s = env.initial_state()
while not env.is_terminal(s):
s = env.next_state(s, np.random.randint(4))
steps += 1
assert steps < 20000