-
Notifications
You must be signed in to change notification settings - Fork 41
/
Maze.py
261 lines (227 loc) · 9.89 KB
/
Maze.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
import numpy as np
import random
import imageio
import matplotlib.pyplot as plt
from os.path import join
class Maze(object):
"""
Maze objects have several main attributes:
- maze_data: wall conditions in every cells are coded as a 4-bit number,
with a bit value taking 0 if there is a wall and 1 if there is no wall.
The 1s register corresponds with a square's top edge, 2s register the
right edge, 4s register the bottom edge, and 8s register the left edge.
"""
valid_actions = ['u', 'r', 'd', 'l'] # Up, Right, Down, Left
direction_bit_map = {'u':1, 'r':2, 'd':4, 'l':8}
# e.g., If there is an opening in the upside of the cell, the cell number & 1 is True
move_map = {
'u': (-1,0),
'r': (0,+1),
'd': (+1,0),
'l': (0,-1),
}
action_unstability = {
'u': {'l':0.1, 'u':0.8, 'r':0.1},
'r': {'u':0.1, 'r':0.8, 'd':0.1},
'd': {'r':0.1, 'd':0.8, 'l':0.1},
'l': {'d':0.1, 'l':0.8, 'u':0.1},
}
robot_img = {d:imageio.imread(join("images/","robot-"+d+".jpg")) for d in valid_actions}
def __init__(self, from_file=None, maze_size=None, trap_number=5, unstable_action=False):
"""
You can construct a map from given file or just generating a random one.
"""
if (from_file is not None) and (maze_size is None):
with open(from_file, 'rb') as f_in:
self.maze_data = np.genfromtxt(from_file,
delimiter=',', dtype=np.uint16)
# Check if the maze have inconsistency in some parts
self.__validate_maze()
elif maze_size is not None:
self.__generate_maze(maze_size[0]*2+1, maze_size[1]*2+1)
else:
raise InputError("Invalid Input")
self.height, self.width = self.maze_data.shape
self.unstable_action = unstable_action
# Generate trap and destination point of the maze
self.__set_destination() # Only one destination
self.__generate_trap(trap_number=trap_number) # Multiple traps
self.__draw_raw_maze_img()
self.__default_robot_loc = {
'loc': (0,self.width-1),
'dir': 'd',
} # Default direction is down
self.place_robot()
self.set_reward()
def __generate_maze(self, height=21, width=27, complexity=.25, density=.25):
"""
Generate a random maze, based on:
https://en.wikipedia.org/wiki/Maze_generation_algorithm
"""
# Only odd shapes
shape = ((height // 2) * 2 + 1, (width // 2) * 2 + 1)
# Adjust complexity and density relative to maze size
complexity = int(complexity * (5 * (shape[0] + shape[1])))
density = int(density * ((shape[0] // 2) * (shape[1] // 2)))
# Build actual maze
Z = np.zeros(shape, dtype=bool)
# Fill borders
Z[0, :] = Z[-1, :] = 1
Z[:, 0] = Z[:, -1] = 1
# Make aisles
for i in range(density):
x, y = random.randint(0, shape[1] // 2) * 2, random.randint(0, shape[0] // 2) * 2
Z[y, x] = 1
for j in range(complexity):
neighbours = []
if x > 1: neighbours.append((y, x - 2))
if x < shape[1] - 2: neighbours.append((y, x + 2))
if y > 1: neighbours.append((y - 2, x))
if y < shape[0] - 2: neighbours.append((y + 2, x))
if len(neighbours):
y_,x_ = neighbours[random.randint(0, len(neighbours) - 1)]
if Z[y_, x_] == 0:
Z[y_, x_] = 1
Z[y_ + (y - y_) // 2, x_ + (x - x_) // 2] = 1
x, y = x_, y_
r,c = Z.shape
# Convert to our maze style
maze_data = np.zeros(((r-3)//2+1,(c-3)//2+1),dtype=np.uint8)
for i in range(0,r-2,2):
for j in range(0,c-2,2):
maze_data[i//2,j//2] = sum([1,2,4,8][i] * ~block for i,block in enumerate(np.ravel(Z[i:i+3,j:j+3],order='F')[[3,7,5,1]]))
self.maze_data = maze_data
def __validate_maze(self):
"""
Check if the input wall contains inconsistency
"""
wall_errors = []
height, width = self.maze_data.shape
# Maze Size Check
if height<=4 or width<=4:
raise InputError("Input maze is too small")
# Vertically Check
for r in range(height-1):
for c in range(width):
if (self.maze_data[r,c] & 4 != 0) != (self.maze_data[r+1,c] & 1 != 0):
wall_errors.append([(r,c), 'v'])
# Horizontally Check
for r in range(height):
for c in range(width-1):
if (self.maze_data[r,c] & 2 != 0) != (self.maze_data[r,c+1] & 8 != 0):
wall_errors.append([(r,c), 'h'])
# Output Errors
if wall_errors:
for cell, wall_type in wall_errors:
if wall_type == 'v':
cell2 = (cell[0]+1, cell[1])
print('Inconsistent vertical wall betweeen {} and {}'.format(cell, cell2))
else:
cell2 = (cell[0], cell[1]+1)
print('Inconsistent horizontal wall betweeen {} and {}'.format(cell, cell2))
raise Exception('Consistency errors found in wall specifications!')
def __set_destination(self, destination_coord=None):
"""
Set destination coordinates, default in center
"""
if not destination_coord:
destination_coord = (self.height//2,self.width//2)
self.destination = destination_coord
def __generate_trap(self, trap_number=5):
"""
Randomly generate traps
"""
if trap_number > self.width * self.height*0.1:
raise ValueError('Too many traps for such small maze')
# Avoid repeated traps
destination = int(self.destination[0] * self.width + self.destination[1])
valid_range = list(range(1,destination)) + list(range(destination+1,int((self.width-1)*(self.height-1))))
trap_list = random.sample(valid_range,trap_number)
self.__traps = [(ele//self.width, ele%self.width) for ele in trap_list]
def __draw_raw_maze_img(self):
# Load grid images
grid_images = []
for i in range(16):
grid_images.append(imageio.imread(join("images/",str(i)+".jpg")))
maze = np.vstack((np.hstack((grid_images[i] for i in row)) for row in self.maze_data))
# Display traps and destination
trap_img = imageio.imread(join("images","trap.jpg"))
dest_img = imageio.imread(join("images","destination.jpg"))
grid_size = 100 # default sizes for grid, trap and destination are 100
for (r,c) in self.__traps:
maze[r*grid_size:(r+1)*grid_size, c*grid_size:(c+1)*grid_size, :] += trap_img
r,c = self.destination
maze[r*grid_size:(r+1)*grid_size, c*grid_size:(c+1)*grid_size, :] += dest_img
# Final maze image
self.__raw_maze_img = maze
def get_raw_maze_img(self):
return self.__raw_maze_img.copy()
def draw_current_maze(self):
grid_size = 100 # default sizes for grid, trap and destination are 100
logo_size = 200 # default sizes for logo is 200
r,c = self.robot['loc']
current_maze_img = self.__raw_maze_img.copy()
current_maze_img[r*grid_size:(r+1)*grid_size, c*grid_size:(c+1)*grid_size, :] += \
self.robot_img[self.robot['dir']]
return current_maze_img
def __repr__(self):
plt.figure(figsize=(self.height,self.width))
plt.imshow(self.draw_current_maze())
plt.axis('off')
plt.show()
return 'Maze of size (%d, %d)'%(self.height, self.width)
def is_permissible(self, location, direction):
"""
Returns a boolean designating whether or not a cell is passable in the
given direction. Cell is input as a tuple. Directions is input as single
letter 'u', 'r', 'd', 'l'.
"""
try:
return (self.maze_data[location] & self.direction_bit_map[direction])!=0
except:
print('Invalid direction or location provided!')
def place_robot(self, robot_loc=None):
"""
Place robot into the maze, default in (0,0)
"""
if not robot_loc:
robot_loc = self.__default_robot_loc.copy()
self.robot = robot_loc
def set_reward(self):
"""
Set rewards for different situations.
"""
self.reward = {
"hit_wall": -10.,
"destination": 50.,
"trap": -30.,
"default": -0.1,
}
def move_robot(self, direction):
"""
Move the robot location according to its location and direction
Return the new location and moving reward
"""
# Random choose action due to action unstability
if not direction in self.valid_actions:
raise ValueError("Invalid Actions")
if self.unstable_action:
unstable_act = self.action_unstability[direction]
direction = np.random.choice(unstable_act.keys(), p=unstable_act.values())
if self.is_permissible(self.robot['loc'],direction):
self.robot['loc'] = tuple((i+di for i,di in zip(self.robot['loc'],self.move_map[direction])))
self.robot['dir'] = direction
if self.robot['loc'] == self.destination:
reward = self.reward['destination']
elif self.robot['loc'] in self.__traps:
reward = self.reward['trap']
else:
reward = self.reward['default']
else:
self.robot['dir'] = direction
reward = self.reward['hit_wall']
return reward
def sense_robot(self):
return self.robot['loc']
def reset_robot(self):
self.robot = self.__default_robot_loc.copy()