def test_bounds_velocity():
env = MountainCar()
env.v = env.BOUND_MIN_V
env.step(0)
assert env.v == env.BOUND_MIN_V
env = MountainCar()
env.v = env.BOUND_MAX_V
env.step(2)
assert env.v == env.BOUND_MAX_V
def test_prepares_observation_space():
env = MountainCar()
assert env.obs_spec[0].shape == (1, )
assert env.obs_spec[0].lo == -1.2
assert env.obs_spec[0].hi == 0.5
assert env.obs_spec[1].shape == (1, )
assert env.obs_spec[1].lo == -0.07
assert env.obs_spec[1].hi == 0.07
assert env.obs_spec[0].is_continuous()
assert env.obs_spec[1].is_continuous()
def test_runs_in_mountain_car_environment():
environment = MountainCar()
agent_program = SemiGradientSarsa(
environment.act_spec,
environment.obs_spec,
alpha=0.5,
epsilon=0.0,
gamma=1.0)
agent = core.Agent(
agent_program,
lambda env: (env.obs, env.reward, env.done()),
lambda action, env: env.step(action))
core.Run(agent, environment).start()
def test_true_sarsa_lambda_runs_in_mountain_car_environment():
environment = MountainCar()
agent_program = TrueOnlineSarsaLambda(
environment.act_spec,
environment.obs_spec,
alpha=0.2,
epsilon=0.0,
gamma=1.0,
lmbda=0.9)
agent = core.Agent(
agent_program,
lambda env: (env.obs, env.reward, env.done()),
lambda action, env: env.step(action))
core.Run(agent, environment).start()
def test_bounds_position():
env = MountainCar()
env.p = -1.19
env.v = env.BOUND_MIN_V
env.step(0)
assert env.p == env.BOUND_MIN_P
assert env.v == 0.0
assert env.reward == -1
def test_actor_critic_runs_in_mountain_car_environment():
environment = MountainCar()
agent_program = ActorCriticLambda(
environment.act_spec,
environment.obs_spec,
alpha_w=0.1,
alpha_theta=0.01,
gamma=1.0,
lambda_w=0.9,
lambda_theta=0.9)
agent = core.Agent(
agent_program,
lambda env: (env.obs, env.reward, env.done()),
lambda action, env: env.step(action))
core.Run(agent, environment).start()
def test_prepares_action_space():
env = MountainCar()
assert env.act_spec[0].shape == (1, )
assert env.act_spec[0].lo == -1
assert env.act_spec[0].hi == 1
def test_is_done_when_top_of_the_mountain_is_reached():
env = MountainCar()
env.p = 0.5
assert env.done()