def prisoners_dilemma():
"""Returns prisoners dilemma game from examples.
Used to test complex behaviours like transitions, events info, etc.
"""
art = ['####d####',
'a A B b',
'#########'
]
item_a = game.Item(color=(0, 254, 254))
item_b = game.Item(color=(254, 254, 0))
item_d = game.Item(color=(0, 254, 254))
items = {'a': item_a, 'b': item_b, 'd': item_d}
player_a = game.Player(color=(0, 100, 254))
player_b = game.Player(color=(254, 100, 0))
players = {'A': player_a, 'B': player_b}
env = game.Game(art, items, players, tabular=True)
env.add_reward('A_moves', {'A': -1})
env.add_reward('B_moves', {'B': -1})
env.add_reward('A_collects_a', {'A': 100})
env.add_reward('B_collects_b', {'B': 100})
env.add_reward('A_collects_d', {'A': 100})
env.add_reward('B_collects_d', {'B': 100})
env.add_terminaison('A_collects_d')
env.add_terminaison('B_collects_d')
env.add_terminaison('A_collects_a')
env.add_terminaison('B_collects_b')
return env
def create_game():
"""Create the assymetric game."""
art = ['############', '#a b ABc #', '############']
item_a = game.Item(color=(0, 254, 254))
item_b = game.Item(color=(254, 254, 0))
item_c = game.Item(color=(0, 254, 254))
items = {'a': item_a, 'b': item_b, 'c': item_c}
player_a = game.Player(color=(0, 100, 254))
player_b = game.Player(color=(254, 100, 0))
players = {'A': player_a, 'B': player_b}
env = game.Game(art, items, players, tabular=True)
env.display()
env.add_reward('A_moves', {'A': -10})
env.add_reward('B_moves', {'B': -10})
env.add_reward('A_collects_a', {'A': 100})
env.add_reward('A_collects_c', {'A': 100})
env.add_reward('B_collects_b', {'B': 100})
env.add_terminaison('A_collects_a')
env.add_terminaison('A_collects_c')
env.add_terminaison('B_collects_b')
# For frame-by-frame visualization:
env = visualizer.Visualizer(env, fps=2, by_episode=False)
# For fast visualization:
# env = visualizer.Visualizer(env, fps=1000, by_episode=True)
return env
def create_game():
"""Creates the temptation game."""
art = [
'########', 'a AB a', 'b b', 'c c', 'd d', 'e e',
'f f', 'g g', 'h h', 'i i', 'j j', '########'
]
item_a = game.Item(color=(60, 254, 254))
item_b = game.Item(color=(80, 254, 254))
item_c = game.Item(color=(100, 254, 254))
item_d = game.Item(color=(120, 254, 254))
item_e = game.Item(color=(140, 254, 254))
item_f = game.Item(color=(160, 254, 254))
item_g = game.Item(color=(180, 254, 254))
item_h = game.Item(color=(200, 254, 254))
item_i = game.Item(color=(220, 254, 254))
item_j = game.Item(color=(254, 254, 254))
items = {
'a': item_a,
'b': item_b,
'c': item_c,
'd': item_d,
'e': item_e,
'f': item_f,
'g': item_g,
'h': item_h,
'i': item_i,
'j': item_j,
}
player_a = game.Player(color=(0, 100, 254))
player_b = game.Player(color=(254, 100, 0))
players = {'A': player_a, 'B': player_b}
env = game.Game(art, items, players, tabular=True)
env.display()
env.add_reward('A_moves', {'A': -1})
env.add_reward('B_moves', {'B': -1})
for i, item in enumerate('abcdefghij'):
env.add_reward('A_collects_' + item, {'A': (i + 1) * 10})
env.add_reward('B_collects_' + item, {'B': (i + 1) * 10})
env.add_terminaison('A_collects_' + item)
env.add_terminaison('B_collects_' + item)
# for frame-by-frame visualization:
env = visualizer.Visualizer(env, fps=2, by_episode=False)
# for fast visualization:
# env = visualizer.Visualizer(env, fps=1000, by_episode=True)
return env
def small_grid():
"""Returns a small grid game, used to test state generation."""
art = ['###', 'A##', '##B']
items = {}
player_a = game.Player(color=(0, 100, 254))
player_b = game.Player(color=(254, 100, 0))
players = {'A': player_a, 'B': player_b}
env = game.Game(art, items, players, tabular=True, max_steps=50)
return env
def test_transition(self):
# (this is the prisonners dilemma grid game from examples)
art = ['####d####', 'a A B b', '#########']
item_a = game.Item(color=(0, 254, 254))
item_b = game.Item(color=(254, 254, 0))
item_d = game.Item(color=(254, 254, 254))
items = {'a': item_a, 'b': item_b, 'd': item_d}
player_a = game.Player(color=(0, 100, 254))
player_b = game.Player(color=(254, 100, 0))
players = {'A': player_a, 'B': player_b}
env = game.Game(art, items, players, tabular=True, max_steps=50)
env.add_reward('A_moves', {'A': -1})
env.add_reward('B_moves', {'B': -1})
env.add_reward('A_collects_a', {'A': 100})
env.add_reward('B_collects_b', {'B': 100})
env.add_reward('A_collects_d', {'A': 100})
env.add_reward('B_collects_d', {'B': 100})
env.add_terminaison('A_collects_d')
env.add_terminaison('B_collects_d')
env.add_terminaison('A_collects_a')
env.add_terminaison('B_collects_b')
_ = env.reset()
# A goes in wall
actions = [1, 0]
_, rewards, _, infos = env.step(actions)
self.assertIn('A_moves', infos['event_list'])
self.assertIn('A_goes_in_walls', infos['event_list'])
self.assertEqual(rewards[0], -1)
_ = env.reset()
# A goes in B
actions = [4, 0]
_ = env.step(actions)
actions = [4, 0]
_, rewards, _, infos = env.step(actions)
self.assertIn('A_moves', infos['event_list'])
self.assertIn('A_blocked_by_B', infos['event_list'])
self.assertEqual(rewards[0], -1)
_ = env.reset()
# A and B reach same cell
actions = [4, 3]
_, rewards, _, infos = env.step(actions)
self.assertIn('A_moves', infos['event_list'])
self.assertIn('B_moves', infos['event_list'])
self.assertTrue(('A_lost_the_drawn' in infos['event_list'])
or ('B_lost_the_drawn' in infos['event_list']))
self.assertEqual(rewards[0], -1)
self.assertEqual(rewards[1], -1)
_ = env.reset()
# A and B block each other
actions = [4, 0]
_ = env.step(actions)
actions = [4, 3]
_, rewards, _, infos = env.step(actions)
self.assertIn('A_moves', infos['event_list'])
self.assertIn('B_moves', infos['event_list'])
self.assertIn('A_blocked_by_B', infos['event_list'])
self.assertIn('B_blocked_by_A', infos['event_list'])
self.assertEqual(rewards[0], -1)
self.assertEqual(rewards[1], -1)
_ = env.reset()
# A reaches reward
actions = [3, 0]
_ = env.step(actions)
actions = [3, 0]
_ = env.step(actions)
actions = [3, 0]
_, rewards, _, infos = env.step(actions)
self.assertIn('A_moves', infos['event_list'])
self.assertIn('A_collects_a', infos['event_list'])
self.assertEqual(rewards[0], 100 - 1)
_ = env.reset()
