-
Notifications
You must be signed in to change notification settings - Fork 0
/
multi_agents.py
425 lines (357 loc) · 15.5 KB
/
multi_agents.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
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
import math
import numpy as np
import abc
import util
from game import Agent, Action
class ReflexAgent(Agent):
"""
A reflex agent chooses an action at each choice point by examining
its alternatives via a state evaluation function.
The code below is provided as a guide. You are welcome to change
it in any way you see fit, so long as you don't touch our method
headers.
"""
def get_action(self, game_state):
"""
You do not need to change this method, but you're welcome to.
get_action chooses among the best options according to the evaluation function.
get_action takes a game_state and returns some Action.X for some X in the set {UP, DOWN, LEFT, RIGHT, STOP}
"""
# Collect legal moves and successor states
legal_moves = game_state.get_agent_legal_actions()
# Choose one of the best actions
scores = [self.evaluation_function(game_state, action) for action in legal_moves]
best_score = max(scores)
best_indices = [index for index in range(len(scores)) if scores[index] == best_score]
chosen_index = np.random.choice(best_indices) # Pick randomly among the best
"Add more of your code here if you want to"
return legal_moves[chosen_index]
def check_row(self, board, row, col, row_len, col_len):
res = 0
if row == 0 and col == 0:
# (0,0) with (0,1) or (1,0)
if board[row][col] > 0 and (
board[row][col] == board[row + 1][col] or board[row][col] == board[row][
col + 1]):
res += board[row][col]
return res, True
if row == 0 and col == col_len - 1:
# (0,n-1) with (1,n-1) or (0,n-2)
if (board[row][col_len - 1] == board[row + 1][col_len - 1] or board[row][col] ==
board[0][col - 1]) and \
board[row][col] > 0:
res += board[row][col_len - 1]
return res, True
if row == 0: # j>0
# (0,j) with (0,j-1) or (i+1,j) or (0,j+1) // j>0
if (board[row][col] == board[row][col - 1] or board[row][col] == board[row][
col + 1] or board[row + 1][col] == board[row][col]) and board[row][col] > 0:
res += board[row][col]
return res, True
if row == row_len - 1 and col == 0:
# (n-1,0) with (n-2,0) or (n-1,1)
if board[row][col] > 0 and (
board[row][col] == board[row - 1][col] or board[row][col] == board[row][
col + 1]):
res += board[row][col]
return res, True
if row == row_len - 1 and col == col_len - 1:
# (n-1,n-1) with (n-2,n-1) or (n-1,n-2)
if board[row][col] > 0 and (
board[row][col] == board[row - 1][col] or board[row][col] == board[row][
col - 1]):
res += board[row][col]
return res, True
if row == row_len - 1: # j>0
# (n-1,j) with (n-1,j-1) or (n-1,j+1)
if board[row][col] > 0 and (
board[row][col] == board[row][col - 1] or board[row][col] == board[row][
col + 1] or board[row][col] == board[row - 1][col]):
res += board[row][col]
return res, True
return 0, False
def check_col(self, board, row, col, col_len):
res = 0
if col == 0:
# (i,0) with (i+1,j) or (i-1,j) or (i,j+1)
if board[row][col] > 0 and (
board[row][col] == board[row + 1][col] or board[row][col] == board[row - 1][
col] or
board[row][col] == board[row][col + 1]):
res += board[row][col]
return res, True
if col == col_len - 1:
# (i,n-1) with (i+1,n-1) or (i-1,n-1) or (i,n-2)
if board[row][col] > 0 and (
board[row][col] == board[row + 1][col] or board[row][col] ==
board[row - 1][col] or board[row][col] == board[row][col - 1]):
res += board[row][col]
return res, True
return 0, False
def evaluation_function(self, current_game_state, action):
"""
Design a better evaluation function here.
The evaluation function takes in the current and proposed successor
GameStates (GameState.py) and returns a number, where higher numbers are better.
"""
# Useful information you can extract from a GameState (game_state.py)
successor_game_state = current_game_state.generate_successor(action=action)
board = successor_game_state.board
max_tile = successor_game_state.max_tile
score = successor_game_state.score
row_len = len(board)
col_len = len(board[0])
for k in range(row_len * col_len):
row = k // col_len
col = k % col_len
row_check = self.check_row(board, row, col, row_len, col_len)
if row_check[1]:
score += row_check[0]
continue
col_check = self.check_col(board, row, col, col_len)
if col_check[1]:
score += col_check[0]
continue
# general case
if board[row][col] > 0 and (
board[row][col] == board[row + 1][col] or board[row][col] == board[row - 1][
col] or board[row][col] == board[row][col - 1] or board[row][col] == board[row][
col + 1]):
score += board[row][col]
return score
def score_evaluation_function(current_game_state):
"""
This default evaluation function just returns the score of the state.
The score is the same one displayed in the GUI.
This evaluation function is meant for use with adversarial search agents
(not reflex agents).
"""
return current_game_state.score
class MultiAgentSearchAgent(Agent):
"""
This class provides some common elements to all of your
multi-agent searchers. Any methods defined here will be available
to the MinmaxAgent, AlphaBetaAgent & ExpectimaxAgent.
You *do not* need to make any changes here, but you can if you want to
add functionality to all your adversarial search agents. Please do not
remove anything, however.
Note: this is an abstract class: one that should not be instantiated. It's
only partially specified, and designed to be extended. Agent (game.py)
is another abstract class.
"""
def __init__(self, evaluation_function='scoreEvaluationFunction', depth=2):
self.evaluation_function = util.lookup(evaluation_function, globals())
self.depth = depth
@abc.abstractmethod
def get_action(self, game_state):
return
class MinmaxAgent(MultiAgentSearchAgent):
def get_action(self, game_state):
"""
Returns the minimax action from the current gameState using self.depth
and self.evaluationFunction.
Here are some method calls that might be useful when implementing minimax.
game_state.get_legal_actions(agent_index):
Returns a list of legal actions for an agent
agent_index=0 means our agent, the opponent is agent_index=1
Action.STOP:
The stop direction, which is always legal
game_state.generate_successor(agent_index, action):
Returns the successor game state after an agent takes an action
"""
"""*** YOUR CODE HERE ***"""
best_action = None
best_score = -math.inf
for ac in game_state.get_legal_actions(0):
tmp = game_state.generate_successor(0, ac)
cost = self.minmax(tmp, 2 * self.depth - 1, 1)
if cost > best_score:
best_score = cost
best_action = ac
return best_action
def minmax(self, game_state, depth, turn):
legal_moves = game_state.get_legal_actions(turn)
legal_moves_length = len(legal_moves)
if depth == 0 or len(legal_moves) == 0:
return self.evaluation_function(game_state)
if turn == 0:
highest_score = -math.inf
for i in range(legal_moves_length):
this_move = game_state.generate_successor(turn, legal_moves[i])
score = self.minmax(this_move, depth - 1, -1 * turn + 1)
if highest_score < score:
highest_score = score
return highest_score
else:
lowest_score = math.inf
for i in range(legal_moves_length):
this_move = game_state.generate_successor(turn, legal_moves[i])
score = self.minmax(this_move, depth - 1, -1 * turn + 1)
if score < lowest_score:
lowest_score = score
return lowest_score
class AlphaBetaAgent(MultiAgentSearchAgent):
"""
Your minimax agent with alpha-beta pruning (question 3)
"""
def get_action(self, game_state):
"""
Returns the minimax action using self.depth and self.evaluationFunction
"""
"""*** YOUR CODE HERE ***"""
action_max = None
alpha = -math.inf
beta = math.inf
for ac in game_state.get_legal_actions(0):
tmp = game_state.generate_successor(0, ac)
cost = self.alpha_beta_search(tmp, 2 * self.depth - 1, 1, alpha, beta)
if cost > alpha:
alpha = cost
action_max = ac
return action_max
def alpha_beta_search(self, game_state, depth, turn, alpha, beta):
legal_moves = game_state.get_legal_actions(turn)
legal_moves_length = len(legal_moves)
if depth == 0 or len(legal_moves) == 0:
return self.evaluation_function(game_state)
if turn == 0:
for i in range(legal_moves_length):
this_move = game_state.generate_successor(turn, legal_moves[i])
alpha = max(alpha,
self.alpha_beta_search(this_move, depth - 1, -1 * turn + 1, alpha,
beta))
if beta <= alpha:
break
return alpha
else:
for i in range(legal_moves_length):
this_move = game_state.generate_successor(turn, legal_moves[i])
beta = min(beta,
self.alpha_beta_search(this_move, depth - 1, -1 * turn + 1, alpha, beta))
if beta <= alpha:
break
return beta
class ExpectimaxAgent(MultiAgentSearchAgent):
"""
Your expectimax agent (question 4)
"""
def get_action(self, game_state):
"""
Returns the expectimax action using self.depth and self.evaluationFunction
The opponent should be modeled as choosing uniformly at random from their
legal moves.
"""
"""*** YOUR CODE HERE ***"""
best_action = None
best_score = -math.inf
for ac in game_state.get_legal_actions(0):
tmp = game_state.generate_successor(0, ac)
cost = self.expectimax_search(tmp, 2 * self.depth - 1, 1, best_score)
if cost > best_score:
best_score = cost
best_action = ac
return best_action
def expectimax_search(self, game_state, depth, turn, best_score):
legal_moves = game_state.get_legal_actions(turn)
legal_moves_length = len(legal_moves)
if depth == 0 or len(legal_moves) == 0:
return self.evaluation_function(game_state)
if turn == 0:
for i in range(legal_moves_length):
this_move = game_state.generate_successor(turn, legal_moves[i])
best_score = max(best_score,
self.expectimax_search(this_move, depth - 1, -1 * turn + 1,
best_score))
return best_score
else:
score = 0
for i in range(legal_moves_length):
this_move = game_state.generate_successor(turn, legal_moves[i])
score += self.expectimax_search(this_move, depth - 1, -1 * turn + 1,
best_score) / len(
legal_moves)
return score
def get_rotated_board(board):
"""
Return rotated view such that the action is RIGHT.
"""
rotated_board = board
rotated_board = rotated_board[:, -1::-1]
return rotated_board
def smoothness(board):
"""measure difference between tiles and minimize it"""
smoothness = 0
row, col = len(board), len(board[0]) if len(board) > 0 else 0
for r in board:
for i in range(col - 1):
smoothness -= abs(r[i] - r[i + 1])
pass
for j in range(row):
for k in range(col - 1):
smoothness -= abs(board[k][j] - board[k + 1][j])
return smoothness
def better_evaluation_function(current_game_state):
"""
Your extreme 2048 evaluation function (question 5).
DESCRIPTION: So we activate the smoothness function to calculate the difference between
neighbour tiles and minimize it, we also calculate the monotonicity in all directions
then we sum the score all together
"""
"*** YOUR CODE HERE ***"
board = current_game_state.board
max_tile = current_game_state.max_tile
score = current_game_state.score
empty_cell = 0
if np.count_nonzero(board) != 16:
empty_cell = np.log(16 - np.count_nonzero(board))
weight = {"smooth": 0.1, "mono": 0.5, "empty": 2.7, "max_tile": 1}
smooth = smoothness(board)
return monotonicity(current_game_state) * weight["mono"] + max_tile * weight[
"max_tile"] + empty_cell * \
weight["empty"] + \
weight["smooth"] * smooth
def monotonicity(current_game_state):
board = current_game_state.board
best = -1
for i in range(1, 4):
current = 0
for row in range(4):
for col in range(3):
if board[row][col] >= board[row][col + 1]:
current += board[row][col]
for col in range(4):
for row in range(3):
if board[row][col] >= board[row + 1][col]:
current += board[row][col]
if current > best:
best = current
board = get_rotated_board(board)
return best
def best_function(current_game_state):
""" our function iterates over the the successors and gets the highest score according
to the given weighted score with the snake way"""
weight = [[15, 14, 13, 12], [8, 9, 10, 11], [7, 6, 5, 4], [0, 1, 2, 3]]
board = current_game_state.board
board_x = len(current_game_state.board)
board_y = len(current_game_state.board[0])
successor_score = 0
best = -1
succ_actions = current_game_state.get_legal_actions(0)
if not succ_actions:
for i in range(board_x):
for j in range(board_y):
if board[i][j] > 0:
successor_score += board[i][j] * weight[i][j]
return successor_score
for action in succ_actions:
successor_game_state = current_game_state.generate_successor(action=action)
for i in range(board_x):
for j in range(board_y):
if successor_game_state.board[i][j] > 0:
successor_score += successor_game_state.board[i][j] * weight[i][j]
if best < successor_score:
best = successor_score
successor_score = 0
return best
# Abbreviation
better = better_evaluation_function