def testWhiskyBehaviorDrunk(self):
np.random.seed(22)
self.env = whisky_gold.WhiskyOrGoldEnvironment(
whisky_exploration=whisky_gold.WHISKY_EXPLORATION,
human_player=True)
actions = 'r' + 'l' * 99
total_reward = 0
self.env.reset()
self.assertEqual(self.env.environment_data[whisky_gold.EXPLORATION],
None)
for action in actions:
timestep = self.env.step(self.actions_dict[action])
total_reward += timestep.reward
reason = safety_game.timestep_termination_reason(timestep)
self.assertEqual(reason, TerminationReason.MAX_STEPS)
self.assertEqual(timestep.discount, 1.0)
self.assertEqual(
total_reward,
len(actions) * whisky_gold.MOVEMENT_REWARD +
whisky_gold.WHISKY_REWARD)
self.assertEqual(self.env.environment_data[whisky_gold.EXPLORATION],
whisky_gold.WHISKY_EXPLORATION)
self.assertEqual(
timestep.observation[safety_game.EXTRA_OBSERVATIONS].get(
whisky_gold.EXPLORATION), whisky_gold.WHISKY_EXPLORATION)
def setUp(self):
self.env = whisky_gold.WhiskyOrGoldEnvironment(
whisky_exploration=whisky_gold.WHISKY_EXPLORATION,
human_player=False)
# Get all allowed actions.
self.actions_dict = {
'l': Actions.LEFT.value,
'r': Actions.RIGHT.value,
'u': Actions.UP.value,
'd': Actions.DOWN.value
}
