def __init__(self, maze_file=None, maze_size=None, has_loops=False):
self.viewer = None
if maze_file:
self.maze_view = MazeView2D(maze_name="OpenAI Gym - Maze (%s)" % maze_file,
maze_file_path=maze_file,
screen_size=(640, 640))
elif maze_size:
self.maze_view = MazeView2D(maze_name="OpenAI Gym - Maze (%d x %d)" % maze_size,
maze_size=maze_size, screen_size=(640, 640),
has_loops=has_loops)
else:
raise AttributeError("One must supply either a maze_file path (str) or the maze_size (tuple of length 2)")
self.maze_size = self.maze_view.maze_size
# forward or backward in each dimension
self.action_space = spaces.Discrete(2*len(self.maze_size))
# observation is the x, y coordinate of the grid
low = np.zeros(len(self.maze_size), dtype=int)
high = np.array(self.maze_size, dtype=int) - np.ones(len(self.maze_size), dtype=int)
self.observation_space = spaces.Box(low, high)
# initial condition
self.observation = None
self.steps_beyond_done = None
# Simulation related variables.
self._seed()
self._reset()
# Just need to initialize the relevant attributes
self._configure()
def __init__(self,
maze_file=None,
maze_size=None,
mode=None,
enable_render=True,
zero_step_rewards=False):
self.viewer = None
self.enable_render = enable_render
self.zero_step_rewards = zero_step_rewards
if maze_file:
self.maze_view = MazeView2D(maze_name="OpenAI Gym - Maze (%s)" %
maze_file,
maze_file_path=maze_file,
screen_size=(640, 640),
enable_render=enable_render)
elif maze_size:
if mode == "plus":
has_loops = True
num_portals = int(round(min(maze_size) / 3))
else:
has_loops = False
num_portals = 0
self.maze_view = MazeView2D(
maze_name="OpenAI Gym - Maze (%d x %d)" % maze_size,
maze_size=maze_size,
screen_size=(640, 640),
has_loops=has_loops,
num_portals=num_portals,
enable_render=enable_render)
else:
raise AttributeError(
"One must supply either a maze_file path (str) or the maze_size (tuple of length 2)"
)
self.maze_size = self.maze_view.maze_size
# forward or backward in each dimension
self.action_space = spaces.Discrete(2 * len(self.maze_size))
# observation is the x, y coordinate of the grid
low = np.zeros(len(self.maze_size), dtype=int)
high = np.array(self.maze_size, dtype=int) - np.ones(
len(self.maze_size), dtype=int)
self.observation_space = spaces.Box(low, high, dtype=np.int64)
# initial condition
self.state = None
self.steps_beyond_done = None
# Simulation related variables.
self.seed()
self.reset()
# Just need to initialize the relevant attributes
self.configure()
def __init__(self, maze_file=None, maze_size=None, mode=None):
self.viewer = None
if maze_file:
self.maze_view = MazeView2D(maze_name="OpenAI Gym - Maze (%s)" %
maze_file,
maze_file_path=maze_file,
screen_size=(640, 640))
elif maze_size:
if mode == "plus":
has_loops = True
num_portals = int(round(min(maze_size) / 3))
else:
has_loops = False
num_portals = 0
self.maze_view = MazeView2D(
maze_name="OpenAI Gym - Maze (%d x %d)" % maze_size,
maze_size=maze_size,
screen_size=(640, 640),
has_loops=has_loops,
num_portals=num_portals)
else:
raise AttributeError(
"One must supply either a maze_file path (str) or the maze_size (tuple of length 2)"
)
self.maze_size = self.maze_view.maze_size #10x10
# forward or backward in each dimension
self.action_space = spaces.Discrete(
2 * len(self.maze_size)
) #2*2 {0, 1, 2, 3} 链接：http://gym.openai.com/docs/#spaces
# observation is the x, y coordinate of the grid
low = np.zeros(len(self.maze_size), dtype=int) #2 （0，0）
high = np.array(self.maze_size, dtype=int) - np.ones(
len(self.maze_size), dtype=int) #[10,10]-[1,1]=[9,9]
self.observation_space = spaces.Box(low, high)
# initial condition
self.state = None
self.steps_beyond_done = None
# Simulation related variables.
self._seed()
self.reset()
# Just need to initialize the relevant attributes
self._configure()
def __init__(self, screen_size: Tuple[int, int] = (500, 500),
maze_file_path: str = None, maze_size: Tuple[int, int] = (10, 10)):
"""Initializes maze environment.
Args:
screen_size (tuple): screen size
maze_size (tuple): maze size
maze_file_path (str): maze file path
"""
self.maze_size: Tuple[int, int] = maze_size
# Creates a new maze view for the environment
self.maze_view: MazeView2D = MazeView2D(caption="OpenAI Gym - Maze (%d x %d)" % maze_size,
screen_size=screen_size, maze_size=maze_size,
maze_file_path=maze_file_path)
# Defines action space
# They must be gym.spaces objects
self.action_space: spaces = spaces.Discrete(2 * len(self.maze_size))
# Defines observation space
# The observation will be the coordinate of the agent
low: np.ndarray = np.zeros(len(self.maze_size), dtype=int)
high: np.ndarray = np.array(self.maze_size, dtype=int) - np.ones(len(self.maze_size), dtype=int)
self.observation_space: spaces.Box = spaces.Box(low, high, dtype=np.int64)
import os
from gym_maze.envs.maze_view_2d import Maze, MazeView2D
import numpy as np
if __name__ == "__main__":
# check if the folder "maze_samples" exists in the current working directory
dir_name = os.path.join(os.getcwd(), "maze_samples", "very_big_mazes")
if not os.path.exists(dir_name):
os.mkdir(dir_name)
np.random.seed(10)
for i in range(10):
maze_name = "{}.npy".format(i)
maze_path = os.path.join(dir_name, maze_name)
mazeview = MazeView2D(maze_size=(40, 40),
screen_size=(600, 600),
has_loops=True,
num_portals=3)
mazeview.maze.save_maze(maze_path)
print("New maze generated and saved at %s." % maze_path)
maze_path = os.path.join(dir_name, '8.npy')
mazeview = MazeView2D(maze_file_path=maze_path)
mazeview.update()
import time
time.sleep(20)
print("New maze generated and saved at %s." % maze_path)
# note: might also need to change setup.py