Skip to content

Instantly share code, notes, and snippets.

@nagataka
Last active January 15, 2021 05:33
Show Gist options
  • Save nagataka/3659810a39ce3b4aefca0ab09291b437 to your computer and use it in GitHub Desktop.
Save nagataka/3659810a39ce3b4aefca0ab09291b437 to your computer and use it in GitHub Desktop.
Blocking Maze for OpenAI Gym
# OpenAI gym custom environment mimicking Blocking Maze
# See Sutton and Barto "Reinforcement Learning an Introduction"
# Example 8.2: Blocking Maze
from enum import Enum
import sys
import copy
import gym
from gym import error, spaces, utils
from gym.utils import seeding
MAP1 = [
"+---------+",
"| G|",
"| |",
"| |",
"|wwwwwwww |",
"| |",
"| s |",
"+---------+",
]
MAP2 = [
"+---------+",
"| G|",
"| |",
"| |",
"| wwwwwwww|",
"| |",
"| s |",
"+---------+",
]
class Action(Enum):
UP = 0
DOWN = 1
LEFT = 2
RIGHT = 3
class MazeEnv1(gym.Env):
metadata = {'render.modes': ['human']}
def __init__(self):
print("init")
self.s_loc = (6, 4)
self.a_loc = (6, 4)
self.g_loc = (1, 9)
self.invalid = ['w', '-', '|']
self.action_space = spaces.Discrete(len(self.actions))
self.observation_space = spaces.Discrete(46) # temporary hard coded
if not hasattr(self, 'map'):
self.map = MAP1
def step(self, action):
current_loc = self.a_loc
if action == Action.UP.value:
self.a_loc = (self.a_loc[0]+1, self.a_loc[1])
elif action == Action.DOWN.value:
self.a_loc = (self.a_loc[0]-1, self.a_loc[1])
elif action == Action.LEFT.value:
self.a_loc = (self.a_loc[0], self.a_loc[1]-1)
elif action == Action.RIGHT.value:
self.a_loc = (self.a_loc[0], self.a_loc[1]+1)
else:
print("Invalid action is specified")
raise ValueError
reward = 0
done = False
info =
if self.map[self.a_loc[0]][self.a_loc[1]] in self.invalid:
# if the move is invalid, return it back
self.a_loc = current_loc
if self.map[self.a_loc[0]][self.a_loc[1]] == 'G':
reward = 1
done = True
return self.a_loc, reward, done, info
def reset(self):
print("reset")
self.__init__()
return self.tuple2int(self.a_loc)
def render(self, mode='human'):
outfile = sys.stdout
m = copy.deepcopy(self.map)
row = m[self.a_loc[0]]
row = row[:self.a_loc[1]]+utils.colorize('A', 'red', bold=True)+row[self.a_loc[1]+1:]
m[self.a_loc[0]] = row
outfile.write("\n".join(["".join(row) for row in m])+"\n")
outfile.write("\n")
def close(self):
print("close")
def switch_maze(self):
self.map = MAP2
self.reset()
@property
def row_length(self):
return len(self.map)-2
@property
def column_length(self):
return len(self.gmap[0])-2
@property
def actions(self):
return [Action.UP, Action.DOWN,
Action.LEFT, Action.RIGHT]
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment