def __init__(self):
#init parent class -> environment interface
env.Env.__init__(self)
#dimensions 1x10x10
self.width = 20
self.height = 16
self.depth = 1
#init state, as 1D vector (tensor with size depth*height*width)
self.observation_init()
#4 actions for movements
self.actions_count = 2
self.player_size = 5
self.reset()
self.gui = gl_gui.GLVisualisation()
self.size_ratio = self.width / self.height
self.player_0_score = 0
self.player_1_score = 0
def __init__(self):
#init parent class -> environment interface
env.Env.__init__(self)
#dimensions 1x1x2
self.width = 1
self.height = 1
self.depth = 2
self.time = 1
#init state, as 1D vector (tensor with size depth*height*width)
self.observation_init()
#4 actions for movements
self.actions_count = 3
self.velocity_min = -0.07
self.velocity_max = 0.07
self.position_min = -1.2
self.position_max = 0.6001
self.move_old = 0
self.move_to_top = 0
self.reset()
self.games_count = 0
self.gui = gl_gui.GLVisualisation()
def __init__(self, size=16):
#init parent class -> environment interface
env.Env.__init__(self)
self.width = size
self.height = size
self.depth = 1
self.time = 1
#init state, as 1D vector (tensor with size depth*height*width)
self.observation_init()
#4 actions for movements
self.actions_count = 3
self.player_size = 3
self.player_0_points = 0.0
self.player_1_points = 0.0
self.game_idx = 0
self.reset()
self.gui = gl_gui.GLVisualisation()
self.size_ratio = self.width / self.height
def __init__(self, name):
libs_env.Env.__init__(self)
self.gym_env = gym.make(name)
self.gym_env.reset()
self.width_scale = 3
self.height_scale = 2
self.raw_width = numpy.shape(self.gym_env.observation_space)[0]
self.raw_height = numpy.shape(self.gym_env.observation_space)[1]
self.raw_depth = numpy.shape(self.gym_env.observation_space)[2]
self.time_steps = 4
self.colors_count = 3
self.width = 96
self.height = 96
self.depth = self.colors_count * self.time_steps
self.actions_count = self.gym_env.action_space.n
size = self.colors_count * self.height * self.width
self.state_vec_frame_0 = numpy.zeros(size)
self.state_vec_frame_1 = numpy.zeros(size)
self.state_vec_frame_2 = numpy.zeros(size)
self.state_vec_frame_3 = numpy.zeros(size)
self.observation_init()
self.gui = libs_gl_gui.GLVisualisation()
self.game_number = 0
def __init__(self):
#init parent class -> environment interface
libs_env.Env.__init__(self)
#dimensions
self.width = 16
self.height = 20
self.depth = 3
#init state, as 1D vector (tensor with size depth*height*width)
self.observation = numpy.zeros(self.get_size())
#3 actions for movements
self.actions_count = 3
#player paddle size
self.player_size = 3
#init game
self.reset()
self.__update_state()
self.gui = libs_gl_gui.GLVisualisation()
self.size_ratio = self.width / self.height
self.moves_old = 0
self.game_idx = 0
self.moves_to_win = 1000
def __init__(self):
env.Env.__init__(self)
self.width = 1
self.height = 1
self.depth = 2
self.actions_count = 2
self.observation_init()
self.reset()
self.gui = gl_gui.GLVisualisation()
def __init__(self):
env.Env.__init__(self)
self.width = 1
self.height = 1
self.depth = 2
self.actions_count = 2
self.observation = numpy.zeros(self.get_size())
self.reset()
self.gui = gl_gui.GLVisualisation()
def __init__(self):
libs.libs_env.env.Env.__init__(self)
self.game = gym_super_mario_bros.make('SuperMarioBros-v0')
self.game = JoypadSpace(self.game, SIMPLE_MOVEMENT)
#self.game.set_sound_enabled(True)
self.game.reset()
state, reward, done, info = self.game.step(0)
self.raw_state_shape = numpy.shape(state)
self.reset()
self.actions_count = 7
self.time_steps = 4
self.colors_count = 3
self.width_scale = 4
self.height_scale = 3
self.width_scaled = self.raw_state_shape[1] // self.width_scale
self.height_scaled = self.raw_state_shape[0] // self.height_scale
self.width = self.width_scaled
self.height = self.height_scaled
self.depth = self.colors_count * self.time_steps
self.observation_init()
print("raw state shape ", self.raw_state_shape[1],
self.raw_state_shape[0])
print("scaling factor ", self.width_scale, self.height_scale)
print("rescaled state ", self.width_scaled, self.height_scaled)
print("state ", self.width, self.height)
self.set_no_terminal_state()
self.game_idx = 0
size = self.colors_count * self.height * self.width
self.state_vec_frame_0 = numpy.zeros(size)
self.state_vec_frame_1 = numpy.zeros(size)
self.state_vec_frame_2 = numpy.zeros(size)
self.state_vec_frame_3 = numpy.zeros(size)
self.size_ratio = self.width / self.height
self.gui = libs_gl_gui.GLVisualisation()
self.window_name = "super mario"
def __init__(self, seed=0, convolution_mode=False):
#init parent class -> environment interface
libs_env.Env.__init__(self)
self.map_size = 64
prime = 60013
if convolution_mode:
self.features_count = 4 + (seed * prime) % 8
self.state_size = 8 #state planar size 8x8
else:
self.features_count = 8 + (seed * prime) % 24
self.state_size = 1 #state planar size 1x1
#state dimensions
self.width = self.state_size
self.height = self.state_size
self.depth = self.features_count
#init state, as 1D vector (tensor with size depth*height*width)
self.observation = numpy.zeros(self.get_size())
#4 actions for movements
self.actions_count = 4
#random permutate actions - to prevent hardcoding bots
self.actions_ids = numpy.random.permutation(self.actions_count)
self.agent_position_x = self.map_size // 2
self.agent_position_y = self.map_size // 2
#create feature maps
self.maps = []
for i in range(0, self.features_count):
init_seed_x = (seed + i) * 61057
init_seed_y = (seed + i) * 62141
map = libs_black_box_map.BlackBoxMap(self.map_size, self.map_size,
init_seed_x, init_seed_y)
self.maps.append(map)
#compute reward by averaging and thresholding feature maps
self.rewards = self.__make_rewards()
self.fields_occupation = numpy.zeros((self.map_size, self.map_size))
#gui for visualisation
self.gui = gl_gui.GLVisualisation()
self.time = 0.0
def __init__(self):
env.Env.__init__(self)
self.width = 4
self.height = 4
self.depth = 1
self.time = 1
self.actions_count = 4
self.observation_init()
self.reset()
self.gui = gl_gui.GLVisualisation()
self.max_value = 0.0
self.max_score_log = rysy.Log("2048_max_score.log")
def __init__(self):
libs_env.Env.__init__(self)
self.size = 16
self.max_depth = 16
self.width = self.size
self.height = self.size
self.depth = 2
self.observation_init()
self.actions_count = 2
self.reset()
self.angle = 0.0
self.gui = libs_gl_gui.GLVisualisation()
def __init__(self):
#init parent class -> environment interface
env.Env.__init__(self)
#dimensions 10x9x1
self.width = 9
self.height = 10
self.depth = 1
self.observation_init()
self.actions_count = 6
self.moves_to_win = 200
self.game_moves = 0.0
self.reset()
self.gui = gl_gui.GLVisualisation()
def __init__(self, size=32):
#init parent class -> environment interface
libs_env.Env.__init__(self)
self.time_steps = 4
self.colors_count = 3
#dimensions
self.width = size
self.height = size
self.depth = self.colors_count
self.time = self.time_steps
#init state, as 1D vector (tensor with size depth*height*width)
self.observation_init()
self.actions_count = 2
self.game_screen = numpy.zeros(
(self.colors_count, self.height, self.width))
size = self.colors_count * self.height * self.width
self.state_vec_frame_0 = numpy.zeros(size)
self.state_vec_frame_1 = numpy.zeros(size)
self.state_vec_frame_2 = numpy.zeros(size)
self.state_vec_frame_3 = numpy.zeros(size)
self.iterations = 0
self.game_idx = 0
self.size_ratio = self.width / self.height
self.gui = libs_gl_gui.GLVisualisation()
self.window_name = "game interface"
self.update_state()
def __init__(self):
#init parent class -> EnvCliff
libs_env_cliff.EnvCliff.__init__(self)
self.gui = libs_gl_gui.GLVisualisation()
def __init__(self, mode="basic"):
libs.libs_env.env.Env.__init__(self)
self.game = DoomGame()
self.mode = mode
self.positive_reward_factor = 1.0
self.negative_reward_factor = 1.0
if self.mode == "basic":
self.positive_reward_factor = 0.1
self.negative_reward_factor = 0.1
self.game.load_config("scenarios/basic.cfg")
left = [1, 0, 0]
right = [0, 1, 0]
attack = [0, 0, 1]
self.actions = [left, right, attack]
if self.mode == "health_gathering":
self.positive_reward_factor = 0.02
self.negative_reward_factor = 0.1
self.game.load_config("scenarios/health_gathering.cfg")
turn_left = [1, 0, 0]
turn_right = [0, 1, 0]
move_forward = [0, 0, 1]
self.actions = [turn_left, turn_right, move_forward]
if self.mode == "defend_the_center":
self.positive_reward_factor = 5.0
self.negative_reward_factor = 10.0
self.game.load_config("scenarios/defend_the_center.cfg")
turn_left = [1, 0, 0]
turn_right = [0, 1, 0]
attack = [0, 0, 1]
self.actions = [turn_left, turn_right, attack]
if self.mode == "defend_the_line":
self.positive_reward_factor = 1.0
self.negative_reward_factor = 10.0
self.game.load_config("scenarios/defend_the_line.cfg")
left = [1, 0, 0]
right = [0, 1, 0]
attack = [0, 0, 1]
self.actions = [left, right, attack]
if self.mode == "deadly_corridor":
self.positive_reward_factor = 1.0
self.negative_reward_factor = 0.5
self.game.load_config("scenarios/deadly_corridor.cfg")
move_left = [1, 0, 0, 0, 0, 0, 0]
move_right = [0, 1, 0, 0, 0, 0, 0]
attack = [0, 0, 1, 0, 0, 0, 0]
move_forward = [0, 0, 0, 1, 0, 0, 0]
move_backward = [0, 0, 0, 0, 1, 0, 0]
turn_left = [0, 0, 0, 0, 0, 1, 0]
turn_right = [0, 0, 0, 0, 0, 0, 1]
self.actions = [
move_left, move_right, attack, move_forward, move_backward,
turn_left, turn_right
]
if self.mode == "multi":
self.positive_reward_factor = 0.1
self.negative_reward_factor = 0.1
self.game.load_config("scenarios/multi.cfg")
turn_left = [1, 0, 0, 0, 0, 0, 0, 0, 0]
turn_right = [0, 1, 0, 0, 0, 0, 0, 0, 0]
attack = [0, 0, 1, 0, 0, 0, 0, 0, 0]
move_right = [0, 0, 0, 1, 0, 0, 0, 0, 0]
move_left = [0, 0, 0, 0, 1, 0, 0, 0, 0]
move_forward = [0, 0, 0, 0, 0, 1, 0, 0, 0]
move_backward = [0, 0, 0, 0, 0, 0, 1, 0, 0]
turn_left_right_delta = [0, 0, 0, 0, 0, 0, 0, 1, 0]
move_left_right_delta = [0, 0, 0, 0, 0, 0, 0, 0, 1]
self.actions = [
turn_left, turn_right, attack, move_right, move_left,
move_forward, move_backward, turn_left_right_delta,
move_left_right_delta
]
if self.mode == "deathmatch":
self.positive_reward_factor = 1.0
self.negative_reward_factor = 1.0
self.game.load_config("scenarios/deathmatch.cfg")
attack = [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
strafe = [0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
move_right = [
0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
]
move_left = [
0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
]
move_forward = [
0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
]
move_backward = [
0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
]
turn_left = [
0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
]
turn_right = [
0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
]
select_weapon_1 = [
0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
]
select_weapon_2 = [
0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0
]
select_weapon_3 = [
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0
]
select_weapon_4 = [
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0
]
select_weapon_5 = [
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0
]
select_weapon_6 = [
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0
]
select_next_weapon = [
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0
]
select_prev_weapon = [
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0
]
look_up_down_delta = [
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0
]
turn_left_right_delta = [
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0
]
move_left_right_delta = [
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1
]
self.actions = [
attack, strafe, move_right, move_left, move_forward,
move_backward, turn_left, turn_right, select_weapon_1,
select_weapon_2, select_weapon_3, select_weapon_4,
select_weapon_5, select_weapon_6, select_next_weapon,
select_prev_weapon, look_up_down_delta, turn_left_right_delta,
move_left_right_delta
]
#self.game.add_available_game_variable(GameVariable.KILLCOUNT)
#self.game.add_available_game_variable(GameVariable.DEATHCOUNT)
#self.game.set_sound_enabled(True)
self.game.init()
self.game.new_episode()
self.reset()
self.raw_state_shape = self.game.get_state().screen_buffer.shape
buffer_size = 64
self.time_steps = 8
self.colors_count = 3
self.width = 80
self.height = 80
self.depth = self.colors_count * self.time_steps
self.time = 1
self.width_scale = 5
self.height_scale = 5
self.width_scaled = self.raw_state_shape[2] // self.width_scale
self.height_scaled = self.raw_state_shape[1] // self.height_scale
self.width_crop = self.raw_state_shape[
2] // self.width_scale - self.width
self.height_crop = self.raw_state_shape[
1] // self.height_scale - self.height
print("raw state shape ", self.raw_state_shape[2],
self.raw_state_shape[1])
print("scaling factor ", self.width_scale, self.height_scale)
print("rescaled state ", self.width_scaled, self.height_scaled)
print("crop ", self.width_crop, self.height_crop)
self.actions_count = len(self.actions)
self.observation_init()
self.game_idx = 0
self.frame_buffer = frame_buffer.FrameBuffer(self.width, self.height,
self.colors_count,
buffer_size)
self.size_ratio = self.width / self.height
self.gui = libs_gl_gui.GLVisualisation()
self.window_name = "DOOM - " + self.mode