Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

grid world with value function rendering #1

Open
wants to merge 1 commit into
base: master
Choose a base branch
from
Open

grid world with value function rendering #1

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
291 changes: 291 additions & 0 deletions GridWorld.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,291 @@
from ast import literal_eval
import gym
import numpy as np
from gym import spaces
from gym.utils import seeding

import matplotlib.pyplot as plt

# Defining actions
UP = 0
RIGHT = 1
DOWN = 2
LEFT = 3
STAY = 4

# Define colors
#COLOURS = {0: [1, 1, 1], 1: [0.6, 0.3, 0.0], 3: [0.0, 1.0, 0.0], 10: [0.6, 0, 1]}
COLOURS = {0: [1, 1, 1], 1: [0.0, 0.0, 0.0], 3: [0.0, 0.0, 0.0], 10: [0.0, 0, 0]}

class GridWorld(gym.Env):
metadata = {'render.modes': ['human']}
MAP = "1 1 1 1 1 1 1 1 1 1 1 1 1\n" \
"1 0 0 0 0 0 1 0 0 0 0 0 1\n" \
"1 0 0 0 0 0 1 0 0 0 0 0 1\n" \
"1 0 0 0 0 0 0 0 0 0 0 0 1\n" \
"1 0 0 0 0 0 1 0 0 0 0 0 1\n" \
"1 0 0 0 0 0 1 0 0 0 0 0 1\n" \
"1 1 0 1 1 1 1 0 0 0 0 0 1\n" \
"1 0 0 0 0 0 1 1 1 0 1 1 1\n" \
"1 0 0 0 0 0 1 0 0 0 0 0 1\n" \
"1 0 0 0 0 0 1 0 0 0 0 0 1\n" \
"1 0 0 0 0 0 0 0 0 0 0 0 1\n" \
"1 0 0 0 0 0 1 0 0 0 0 0 1\n" \
"1 1 1 1 1 1 1 1 1 1 1 1 1"

def __init__(self, MAP=MAP, goal_reward=1.0, step_reward=-0.1, goals=None, continual=False, random_start_state=True):

self.n = None
self.m = None

self.grid = None
self.hallwayStates = None
self.possibleStates = []
self.walls = []

self.MAP = MAP
self._map_init()

self.continual = continual

self.random_start_state = random_start_state
self.start_state_coord = (1, 1)
self.state = self.start_state_coord

self.done = False

self.goalReached = None
self.goals = [(1, 1)]
if goals:
self.goals = goals

# Rewards
self.step_reward = step_reward
self.goal_reward = goal_reward
self.hit_wall_reward = step_reward

# Gym spaces for observation and action space
self.observation_space = spaces.Discrete(len(self.possibleStates))
self.action_space = spaces.Discrete(5)

def seed(self, seed=None):
self.np_random, seed = seeding.np_random(seed)
np.random.rand(seed)
return [seed]

def step(self, action):
assert self.action_space.contains(action)

if action == STAY and self.state in self.goals and not self.continual:
self.done = True

reward = self._get_reward(self.state, action)

x, y = self.state
if action == UP:
x = x - 1
elif action == DOWN:
x = x + 1
elif action == RIGHT:
y = y + 1
elif action == LEFT:
y = y - 1
new_state = (x, y)

if self._get_grid_value(new_state) == 1: # new_state in walls list
# stay at old state if new coord is wall
new_state = self.state
else:
self.state = new_state

return [self.goalReached, self.state], reward, self.done, None

def reset(self):
self.done = False
self.goalReached = None
if self.random_start_state:
idx = np.random.randint(len(self.possibleStates))
self.state = self.possibleStates[idx] # self.start_state_coord
else:
self.state = self.start_state_coord
return [self.goalReached, self.state]

def render(self, mode='human', draw_arrows=False, draw_values=False, draw_rewards=False, V = None, policy=None, R = None, title=None, grid=False, cmap='RdBu'):

img = self._gridmap_to_img()
fig = plt.figure(1, figsize=(10, 8), dpi=60, facecolor='w', edgecolor='k')

plt.clf()
plt.xticks(np.arange(0, 2*self.n, 1))
plt.yticks(np.arange(0, 2*self.m, 1))
plt.grid(grid)
if title:
plt.title(title)

plt.imshow(img, origin="upper", extent=[0, self.n, self.m, 0])
fig.canvas.draw()

if draw_rewards & (type(R) is not None): # For showing optimal values
r = np.zeros((self.n,self.m))+float("-inf")
for state, value in R.items():
y, x = literal_eval(state)[1]
r[y,x] = value.sum()
c = plt.imshow(r, origin="upper", cmap=cmap, extent=[0, self.n, self.m, 0])
fig.colorbar(c, ax=fig.gca())

if draw_values & (type(V) is not None): # For showing optimal values
v = np.zeros((self.n,self.m))+float("-inf")
for state, value in V.items():
y, x = literal_eval(state)[1]
v[y,x] = value
c = plt.imshow(v, origin="upper", cmap=cmap, extent=[0, self.n, self.m, 0])
fig.colorbar(c, ax=fig.gca())

if draw_arrows & (type(policy) is not None): # For drawing arrows of optimal policy
fig = plt.gcf()
ax = fig.gca()
for state, action in policy.items():
y, x = literal_eval(state)[1]
self._draw_arrows(fig, ax, x, y, action)

plt.pause(0.00001) # 0.01
return

def _map_init(self):
self.grid = []
lines = self.MAP.split('\n')

for i, row in enumerate(lines):
row = row.split(' ')
if self.n is not None and len(row) != self.n:
raise ValueError(
"Map's rows are not of the same dimension...")
self.n = len(row)
rowArray = []
for j, col in enumerate(row):
rowArray.append(int(col))
if col == "1":
self.walls.append((i, j))
# possible states
else:
self.possibleStates.append((i, j))
self.grid.append(rowArray)
self.m = i + 1

self._find_hallWays()

def _find_hallWays(self):
self.hallwayStates = []
for x, y in self.possibleStates:
if ((self.grid[x - 1][y] == 1) and (self.grid[x + 1][y] == 1)) or \
((self.grid[x][y - 1] == 1) and (self.grid[x][y + 1] == 1)):
self.hallwayStates.append((x, y))

def _get_grid_value(self, state):
return self.grid[state[0]][state[1]]

# specific for self.MAP
def _getRoomNumber(self, state=None):
if state == None:
state = self.state
# if state isn't at hall way point
xCount = self._greaterThanCounter(state, 0)
yCount = self._greaterThanCounter(state, 1)
room = 0
if yCount >= 2:
if xCount >= 2:
room = 2
else:
room = 1
else:
if xCount >= 2:
room = 3
else:
room = 0

return room

def _greaterThanCounter(self, state, index):
count = 0
for h in self.hallwayStates:
if state[index] > h[index]:
count = count + 1
return count

def _get_reward(self, state, action):
reward = 0
if action == STAY and self.state in self.goals and not self.goalReached:
self.goalReached = state
reward = self.goal_reward
elif action == STAY and self.state in self.goals:
reward = self.step_reward/4
elif action == STAY:
reward = self.step_reward/2
elif self._get_grid_value(state) == 1:
reward = self.hit_wall_reward
else:
reward = self.step_reward
return reward

def _draw_arrows(self, fig, ax, x, y, direction):
if direction == UP:
x += 0.5
y += 1
dx = 0
dy = -0.4
if direction == DOWN:
x += 0.5
dx = 0
dy = 0.4
if direction == RIGHT:
y += 0.5
dx = 0.4
dy = 0
if direction == LEFT:
x += 1
y += 0.5
dx = -0.4
dy = 0
if direction == STAY:
x += 0.5
y += 0.5
dx = 0
dy = 0

ax.add_patch(plt.Circle((x, y), radius=0.25, fc='k'))
return

plt.arrow(x, # x1
y, # y1
dx, # x2 - x1
dy, # y2 - y1
facecolor='k',
edgecolor='k',
width=0.005,
head_width=0.4,
)

def _gridmap_to_img(self):
row_size = len(self.grid)
col_size = len(self.grid[0])

obs_shape = [row_size, col_size, 3]

img = np.zeros(obs_shape)

gs0 = int(img.shape[0] / row_size)
gs1 = int(img.shape[1] / col_size)
for i in range(row_size):
for j in range(col_size):
for k in range(3):
if (i, j) == self.state:
this_value = COLOURS[10][k]
elif (i, j) in self.goals:
this_value = COLOURS[3][k]
else:

colour_number = int(self.grid[i][j])
this_value = COLOURS[colour_number][k]
img[i * gs0:(i + 1) * gs0, j * gs1:(j + 1)
* gs1, k] = this_value
return img