def test_get_rba_array():
maze = Map('classic')
g = Game(maze=maze,
screen=pg.display.set_mode(maze.get_map_sizes()),
sounds_active=False,
state_active=False)
arr = g.get_rba_array()
assert type(arr) is np.ndarray
assert arr.shape == (456, 552, 3)
class PacmanEnv(gym.Env):
"""
Reinforcement Learning Environment wrapper for the game.
It encapsulates an environment with arbitrary behind-the-scenes dynamics.
An environment can be partially or fully observed.
The main API methods that users of this class need to know are:
step
reset
render
close
seed
Its extends the gym.Env class
"""
metadata = {
'render.modes': ['human', 'rgb_array'],
'video.frames_per_second': 30
}
reward_range = (-10, 5)
def __init__(self,
layout: str,
enable_render=True,
state_active=False,
player_lives: int = 3):
"""
PacmanEnv constructor
:param layout: the layout of the game
:param frame_to_skip: the frame to skip during training
:param enable_render: enabling the display of the game screen
:param state_active: enabling the display of the state matrix
"""
self.layout = layout
self.state_active = state_active
self.enable_render = enable_render
if enable_render:
pg.init()
self.action_space = spaces.Discrete(Action.__len__())
self.maze = Map(layout)
self.width, self.height = self.maze.get_map_sizes()
self.game = Game(
maze=self.maze,
screen=Controller.get_screen(state_active, self.width, self.height)
if enable_render else None,
sounds_active=False,
state_active=state_active,
agent=None)
self.timer = 0
self.reinit_game = False
self.player_lives = player_lives
self.observation_space = spaces.Space(
shape=self.get_screen_rgb_array().shape, dtype=int)
self.seed()
def __del__(self):
if self.enable_render:
pg.quit()
def seed(self, seed=None):
"""
Sets the seed for this env's random number generator(s).
:param seed:
:return:
"""
_, seed = seeding.np_random(seed)
return [seed]
def reset(self, mode='human'):
"""
Reset and Restart the Pacman Environment
"""
self.game.maze.reinit_map()
self.game.restart()
self.game.player.regenerate()
self.game.score = 0
self.game.mode_timer = 0
self.game.ghosts_timer = 0
self.game.set_mode(GameMode.normal)
self.game.make_ghosts_normal()
self.game.player.lives = self.player_lives
if mode == 'human':
return self.get_state_matrix()
elif mode == 'rgb_array':
return self.get_screen_rgb_array()
elif mode == 'info':
return self.get_info_dict()
def render(self, mode='human'):
"""
Renders the environment.
The set of supported modes varies per environment. (And some
environments do not support rendering at all.) By convention,
if mode is:
- human: render to the current display or terminal and
return nothing. Usually for human consumption.
- rgb_array: Return an numpy.ndarray with shape (x, y, 3),
representing RGB values for an x-by-y pixel image, suitable
for turning into a video.
:param mode: the mode to render
:return: the rba_array if the anonymous mode is active
"""
if mode == 'human':
if self.enable_render:
self.game.init_screen()
self.game.draw()
pg.display.flip()
elif mode == 'rgb_array':
return self.get_screen_rgb_array()
def close(self):
self.__del__()
def step(self, action: Union[Action, int]):
"""
Run one timestep of the environment's dynamics. When end of
episode is reached, you are responsible for calling `reset()`
to reset this environment's state.
Accepts an action and returns a tuple (observation, reward, done, info).
:param action: action to perform in the environment
:return: a tuple containing the following: observation, reward, done, info
"""
action = Action(int(action)) if type(action) is int else action
rewards = self._one_step_action(action)
done = self.get_mode() == GameMode.game_over or self.get_mode(
) == GameMode.black_screen
obs = self.get_screen_rgb_array()
info = self.get_info_dict()
return obs, rewards, done, info
def get_info_dict(self) -> Dict[str, Any]:
ghosts_pixel_pos = [
ghost.get_pixel_position() for ghost in self.game.ghosts
]
number_of_scared_ghosts = sum(
[ghost.is_vulnerable() for ghost in self.game.ghosts])
info = {
'win': self.get_mode() == GameMode.black_screen,
'player position': self.get_player_position(),
'player pixel position': self.get_player_pixel_position(),
'player lives': self.game.player.lives,
'game mode': self.get_mode().value,
'game score': self.game.score,
'number of scared ghosts': number_of_scared_ghosts,
'state matrix': self.get_state_matrix(),
'ghosts_pixel_pos': ghosts_pixel_pos,
'player vel': self.game.player.get_vel(),
'player action': self.game.player.current_action
}
return info
def _one_step_action(self, action: Union[Action, int]) -> int:
"""
Performs only one step of the given action in the environment
:param action: action to perform
:return: the reward obtained after performing the action
"""
self.check_game_mode()
if self.get_mode() is GameMode.game_over:
return 0
elif self.get_mode() is GameMode.black_screen:
return 0
elif self.reinit_game:
self.reinit_game = False
return 0
prev_reward = self.game.total_rewards
self.game.player.change_player_vel(action, self.game)
self.game.move_players()
succ_reward = self.game.total_rewards
return succ_reward - prev_reward
def get_mode(self) -> GameMode:
"""
:return: the current mode of the Game, is an instance of the Enum GameMode
"""
return self.game.game_mode
def get_state_matrix(self) -> np.ndarray:
return self.maze.state_matrix
def get_player_position(self) -> Tuple[int, int]:
"""
:return: a tuple containing respectively the x and y coordinate
int the grid
"""
return self.game.player.nearest_col, self.game.player.nearest_row
def get_player_pixel_position(self) -> Tuple[int, int]:
"""
:return: a tuple containing respectively the x and y pixel position in the
game
"""
return self.game.player.x, self.game.player.y
def check_game_mode(self):
mode = self.get_mode()
if self.maze.get_number_of_pellets() == 0:
self.game.set_mode(GameMode.black_screen)
return
if mode is GameMode.hit_ghost:
self.game.player.lives -= 1
if self.game.player.lives == 0:
self.game.set_mode(GameMode.game_over)
else:
self.game.init_players_in_map()
self.game.make_ghosts_normal()
self.game.set_mode(GameMode.normal)
self.reinit_game = True
elif mode == GameMode.wait_after_eating_ghost:
self.game.move_ghosts()
if self.maze.get_number_of_pellets() == 0:
self.game.set_mode(GameMode.black_screen)
elif self.game.are_all_ghosts_vulnerable():
self.game.set_mode(GameMode.change_ghosts)
elif self.game.are_all_ghosts_normal():
self.game.set_mode(GameMode.normal)
self.game.check_ghosts_state()
def get_screen_rgb_array(self):
screen = self.game.screen.copy()
return pg.surfarray.pixels3d(screen)