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
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
|
- # game.py
- # -------
- # Licensing Information: Please do not distribute or publish solutions to this
- # project. You are free to use and extend these projects for educational
- # purposes. The Pacman AI projects were developed at UC Berkeley, primarily by
- # John DeNero (denero@cs.berkeley.edu) and Dan Klein (klein@cs.berkeley.edu).
- # For more info, see http://inst.eecs.berkeley.edu/~cs188/sp09/pacman.html
- from utilities.util import *
- from utilities.util import raiseNotDefined
- import time
- import traceback
- try:
- import boinc
- _BOINC_ENABLED = True
- except:
- _BOINC_ENABLED = False
- #######################
- # Parts worth reading #
- #######################
- class Agent:
- """
- An agent must define a getAction method, but may also define the
- following methods which will be called if they exist:
- def registerInitialState(self, state): # inspects the starting state
- """
- def __init__(self, index=0):
- self.index = index
- def getAction(self, state):
- """
- The Agent will receive a GameState (from either {pacman, capture, sonar}.py) and
- must return an action from Directions.{North, South, East, West, Stop}
- """
- raiseNotDefined()
- class Directions:
- NORTH = 'North'
- SOUTH = 'South'
- EAST = 'East'
- WEST = 'West'
- STOP = 'Stop'
- LEFT = {NORTH: WEST,
- SOUTH: EAST,
- EAST: NORTH,
- WEST: SOUTH,
- STOP: STOP}
- RIGHT = dict([(y, x) for x, y in LEFT.items()])
- REVERSE = {NORTH: SOUTH,
- SOUTH: NORTH,
- EAST: WEST,
- WEST: EAST,
- STOP: STOP}
- class Configuration:
- """
- A Configuration holds the (x,y) coordinate of a character, along with its
- traveling direction.
- The convention for positions, like a graph, is that (0,0) is the lower left corner, x increases
- horizontally and y increases vertically. Therefore, north is the direction of increasing y, or (0,1).
- """
- def __init__(self, pos, direction):
- self.pos = pos
- self.direction = direction
- def getPosition(self):
- return (self.pos)
- def getDirection(self):
- return self.direction
- def isInteger(self):
- x, y = self.pos
- return x == int(x) and y == int(y)
- def __eq__(self, other):
- if other == None: return False
- return (self.pos == other.pos and self.direction == other.direction)
- def __hash__(self):
- x = hash(self.pos)
- y = hash(self.direction)
- return hash(x + 13 * y)
- def __str__(self):
- return "(x,y)=" + str(self.pos) + ", " + str(self.direction)
- def generateSuccessor(self, vector):
- """
- Generates a new configuration reached by translating the current
- configuration by the action vector. This is a low-level call and does
- not attempt to respect the legality of the movement.
- Actions are movement vectors.
- """
- x, y = self.pos
- dx, dy = vector
- direction = Actions.vectorToDirection(vector)
- if direction == Directions.STOP:
- direction = self.direction # There is no stop direction
- return Configuration((x + dx, y + dy), direction)
- class AgentState:
- """
- AgentStates hold the state of an agent (configuration, speed, scared, etc).
- """
- def __init__(self, startConfiguration, isPacman):
- self.start = startConfiguration
- self.configuration = startConfiguration
- self.isPacman = isPacman
- self.scaredTimer = 0
- def __str__(self):
- if self.isPacman:
- return "Pacman: " + str(self.configuration)
- else:
- return "Ghost: " + str(self.configuration)
- def __eq__(self, other):
- if other == None:
- return False
- return self.configuration == other.configuration and self.scaredTimer == other.scaredTimer
- def __hash__(self):
- return hash(hash(self.configuration) + 13 * hash(self.scaredTimer))
- def copy(self):
- state = AgentState(self.start, self.isPacman)
- state.configuration = self.configuration
- state.scaredTimer = self.scaredTimer
- return state
- def getPosition(self):
- if self.configuration == None: return None
- return self.configuration.getPosition()
- def getDirection(self):
- return self.configuration.getDirection()
- class Grid:
- """
- A 2-dimensional array of objects backed by a list of lists. Data is accessed
- via grid[x][y] where (x,y) are positions on a Pacman map with x horizontal,
- y vertical and the origin (0,0) in the bottom left corner.
- The __str__ method constructs an output that is oriented like a pacman board.
- """
- def __init__(self, width, height, initialValue=False, bitRepresentation=None):
- if initialValue not in [False, True]: raise Exception('Grids can only contain booleans')
- self.CELLS_PER_INT = 30
- self.width = width
- self.height = height
- self.data = [[initialValue for y in range(height)] for x in range(width)]
- self.__class__.__lt__ = lambda x, y: (True)
- if bitRepresentation:
- self._unpackBits(bitRepresentation)
- def __getitem__(self, i):
- return self.data[i]
- def __setitem__(self, key, item):
- self.data[key] = item
- def __str__(self):
- out = [[str(self.data[x][y])[0] for x in range(self.width)] for y in range(self.height)]
- out.reverse()
- return '\n'.join([''.join(x) for x in out])
- def __eq__(self, other):
- if other == None: return False
- return self.data == other.data
- def __hash__(self):
- # return hash(str(self))
- base = 1
- h = 0
- for l in self.data:
- for i in l:
- if i:
- h += base
- base *= 2
- return hash(h)
- def copy(self):
- g = Grid(self.width, self.height)
- g.data = [x[:] for x in self.data]
- return g
- def deepCopy(self):
- return self.copy()
- def shallowCopy(self):
- g = Grid(self.width, self.height)
- g.data = self.data
- return g
- def count(self, item=True):
- return sum([x.count(item) for x in self.data])
- def asList(self, key=True):
- list = []
- for x in range(self.width):
- for y in range(self.height):
- if self[x][y] == key: list.append((x, y))
- return list
- def packBits(self):
- """
- Returns an efficient int list representation
- (width, height, bitPackedInts...)
- """
- bits = [self.width, self.height]
- currentInt = 0
- for i in range(self.height * self.width):
- bit = self.CELLS_PER_INT - (i % self.CELLS_PER_INT) - 1
- x, y = self._cellIndexToPosition(i)
- if self[x][y]:
- currentInt += 2 ** bit
- if (i + 1) % self.CELLS_PER_INT == 0:
- bits.append(currentInt)
- currentInt = 0
- bits.append(currentInt)
- return tuple(bits)
- def _cellIndexToPosition(self, index):
- x = index / self.height
- y = index % self.height
- return x, y
- def _unpackBits(self, bits):
- """
- Fills in data from a bit-level representation
- """
- cell = 0
- for packed in bits:
- for bit in self._unpackInt(packed, self.CELLS_PER_INT):
- if cell == self.width * self.height: break
- x, y = self._cellIndexToPosition(cell)
- self[x][y] = bit
- cell += 1
- def _unpackInt(self, packed, size):
- bools = []
- if packed < 0: raise ValueError("must be a positive integer")
- for i in range(size):
- n = 2 ** (self.CELLS_PER_INT - i - 1)
- if packed >= n:
- bools.append(True)
- packed -= n
- else:
- bools.append(False)
- return bools
- def reconstituteGrid(bitRep):
- if type(bitRep) is not type((1, 2)):
- return bitRep
- width, height = bitRep[:2]
- return Grid(width, height, bitRepresentation=bitRep[2:])
- ####################################
- # Parts you shouldn't have to read #
- ####################################
- class Actions:
- """
- A collection of static methods for manipulating move actions.
- """
- # Directions
- _directions = {Directions.NORTH: (0, 1),
- Directions.SOUTH: (0, -1),
- Directions.EAST: (1, 0),
- Directions.WEST: (-1, 0),
- Directions.STOP: (0, 0)}
- _directionsAsList = _directions.items()
- TOLERANCE = .001
- def reverseDirection(action):
- if action == Directions.NORTH:
- return Directions.SOUTH
- if action == Directions.SOUTH:
- return Directions.NORTH
- if action == Directions.EAST:
- return Directions.WEST
- if action == Directions.WEST:
- return Directions.EAST
- return action
- reverseDirection = staticmethod(reverseDirection)
- def vectorToDirection(vector):
- dx, dy = vector
- if dy > 0:
- return Directions.NORTH
- if dy < 0:
- return Directions.SOUTH
- if dx < 0:
- return Directions.WEST
- if dx > 0:
- return Directions.EAST
- return Directions.STOP
- vectorToDirection = staticmethod(vectorToDirection)
- def directionToVector(direction, speed=1.0):
- dx, dy = Actions._directions[direction]
- return (dx * speed, dy * speed)
- directionToVector = staticmethod(directionToVector)
- def getPossibleActions(config, walls):
- possible = []
- x, y = config.pos
- x_int, y_int = int(x + 0.5), int(y + 0.5)
- # In between grid points, all agents must continue straight
- if (abs(x - x_int) + abs(y - y_int) > Actions.TOLERANCE):
- return [config.getDirection()]
- for dir, vec in Actions._directionsAsList:
- dx, dy = vec
- next_y = y_int + dy
- next_x = x_int + dx
- if not walls[next_x][next_y]: possible.append(dir)
- return possible
- getPossibleActions = staticmethod(getPossibleActions)
- def getLegalNeighbors(position, walls):
- x, y = position
- x_int, y_int = int(x + 0.5), int(y + 0.5)
- neighbors = []
- for dir, vec in Actions._directionsAsList:
- dx, dy = vec
- next_x = x_int + dx
- if next_x < 0 or next_x == walls.width: continue
- next_y = y_int + dy
- if next_y < 0 or next_y == walls.height: continue
- if not walls[next_x][next_y]: neighbors.append((next_x, next_y))
- return neighbors
- getLegalNeighbors = staticmethod(getLegalNeighbors)
- def getSuccessor(position, action):
- dx, dy = Actions.directionToVector(action)
- x, y = position
- return (x + dx, y + dy)
- getSuccessor = staticmethod(getSuccessor)
- class GameStateData:
- def __init__(self, prevState=None):
- """
- Generates a new data packet by copying information from its predecessor.
- """
- if prevState != None:
- self.food = prevState.food.shallowCopy()
- self.capsules = prevState.capsules[:]
- self.agentStates = self.copyAgentStates(prevState.agentStates)
- self.layout = prevState.layout
- self._eaten = prevState._eaten
- self.score = prevState.score
- self._foodEaten = None
- self._capsuleEaten = None
- self._agentMoved = None
- self._lose = False
- self._win = False
- self.scoreChange = 0
- def deepCopy(self):
- state = GameStateData(self)
- state.food = self.food.deepCopy()
- state.layout = self.layout.deepCopy()
- state._agentMoved = self._agentMoved
- state._foodEaten = self._foodEaten
- state._capsuleEaten = self._capsuleEaten
- return state
- def copyAgentStates(self, agentStates):
- copiedStates = []
- for agentState in agentStates:
- copiedStates.append(agentState.copy())
- return copiedStates
- def __eq__(self, other):
- """
- Allows two states to be compared.
- """
- if other == None: return False
- # TODO Check for type of other
- if not self.agentStates == other.agentStates: return False
- if not self.food == other.food: return False
- if not self.capsules == other.capsules: return False
- if not self.score == other.score: return False
- return True
- def __hash__(self):
- """
- Allows states to be keys of dictionaries.
- """
- for i, state in enumerate(self.agentStates):
- try:
- int(hash(state))
- except(TypeError, e):
- print(e)
- # hash(state)
- return int((hash(tuple(self.agentStates)) + 13 * hash(self.food) + 113 * hash(
- tuple(self.capsules)) + 7 * hash(self.score)) % 1048575)
- def __str__(self):
- width, height = self.layout.width, self.layout.height
- map = Grid(width, height)
- if type(self.food) == type((1, 2)):
- self.food = reconstituteGrid(self.food)
- for x in range(width):
- for y in range(height):
- food, walls = self.food, self.layout.walls
- map[x][y] = self._foodWallStr(food[x][y], walls[x][y])
- for agentState in self.agentStates:
- if agentState == None: continue
- if agentState.configuration == None: continue
- x, y = [int(i) for i in nearestPoint(agentState.configuration.pos)]
- agent_dir = agentState.configuration.direction
- if agentState.isPacman:
- map[x][y] = self._pacStr(agent_dir)
- else:
- map[x][y] = self._ghostStr(agent_dir)
- for x, y in self.capsules:
- map[x][y] = 'o'
- return str(map) + ("\nScore: %d\n" % self.score)
- def _foodWallStr(self, hasFood, hasWall):
- if hasFood:
- return '.'
- elif hasWall:
- return '%'
- else:
- return ' '
- def _pacStr(self, dir):
- if dir == Directions.NORTH:
- return 'v'
- if dir == Directions.SOUTH:
- return '^'
- if dir == Directions.WEST:
- return '>'
- return '<'
- def _ghostStr(self, dir):
- return 'G'
- if dir == Directions.NORTH:
- return 'M'
- if dir == Directions.SOUTH:
- return 'W'
- if dir == Directions.WEST:
- return '3'
- return 'E'
- def initialize(self, layout, numGhostAgents):
- """
- Creates an initial game state from a layout array (see layout.py).
- """
- self.food = layout.food.copy()
- self.capsules = layout.capsules[:]
- self.layout = layout
- self.score = 0
- self.scoreChange = 0
- self.agentStates = []
- numGhosts = 0
- for isPacman, pos in layout.agentPositions:
- if not isPacman:
- if numGhosts == numGhostAgents:
- continue # Max ghosts reached already
- else:
- numGhosts += 1
- self.agentStates.append(AgentState(Configuration(pos, Directions.STOP), isPacman))
- self._eaten = [False for a in self.agentStates]
- class Game:
- """
- The Game manages the control flow, soliciting actions from agents.
- """
- def __init__(self, agents, display, rules, startingIndex=0, muteAgents=False,
- catchExceptions=False):
- self.agentCrashed = False
- self.agents = agents
- self.display = display
- self.rules = rules
- self.startingIndex = startingIndex
- self.gameOver = False
- self.muteAgents = muteAgents
- self.catchExceptions = catchExceptions
- self.moveHistory = []
- self.totalAgentTimes = [0 for agent in agents]
- self.totalAgentTimeWarnings = [0 for agent in agents]
- self.agentTimeout = False
- def getProgress(self):
- if self.gameOver:
- return 1.0
- else:
- return self.rules.getProgress(self)
- def _agentCrash(self, agentIndex, quiet=False):
- """
- Helper method for handling agent crashes
- """
- if not quiet: traceback.print_exc()
- self.gameOver = True
- self.agentCrashed = True
- self.rules.agentCrash(self, agentIndex)
- OLD_STDOUT = None
- OLD_STDERR = None
- def mute(self):
- if not self.muteAgents: return
- global OLD_STDOUT, OLD_STDERR
- import cStringIO
- OLD_STDOUT = sys.stdout
- OLD_STDERR = sys.stderr
- sys.stdout = cStringIO.StringIO()
- sys.stderr = cStringIO.StringIO()
- def unmute(self):
- if not self.muteAgents: return
- global OLD_STDOUT, OLD_STDERR
- sys.stdout.close()
- sys.stderr.close()
- # Revert stdout/stderr to originals
- sys.stdout = OLD_STDOUT
- sys.stderr = OLD_STDERR
- def run(self):
- """
- Main control loop for game play.
- """
- self.display.initialize(self.state.data)
- self.numMoves = 0
- ###self.display.initialize(self.state.makeObservation(1).data)
- # inform learning agents of the game start
- for i in range(len(self.agents)):
- agent = self.agents[i]
- if not agent:
- # this is a null agent, meaning it failed to load
- # the other team wins
- self._agentCrash(i, quiet=True)
- return
- if ("registerInitialState" in dir(agent)):
- self.mute()
- if self.catchExceptions:
- try:
- timed_func = TimeoutFunction(agent.registerInitialState,
- int(self.rules.getMaxStartupTime(i)))
- try:
- start_time = time.time()
- timed_func(self.state.deepCopy())
- time_taken = time.time() - start_time
- self.totalAgentTimes[i] += time_taken
- except(TimeoutFunctionException):
- print("Agent %d ran out of time on startup!" % i)
- self.unmute()
- self.agentTimeout = True
- self._agentCrash(i, quiet=True)
- return
- except(Exception):
- self.unmute()
- self._agentCrash(i, quiet=True)
- return
- else:
- agent.registerInitialState(self.state.deepCopy())
- ## TODO: could this exceed the total time
- self.unmute()
- agentIndex = self.startingIndex
- numAgents = len(self.agents)
- while not self.gameOver:
- # Fetch the next agent
- agent = self.agents[agentIndex]
- move_time = 0
- skip_action = False
- # Generate an observation of the state
- if 'observationFunction' in dir(agent):
- self.mute()
- if self.catchExceptions:
- try:
- timed_func = TimeoutFunction(agent.observationFunction,
- int(self.rules.getMoveTimeout(agentIndex)))
- try:
- start_time = time.time()
- observation = timed_func(self.state.deepCopy())
- except(TimeoutFunctionException):
- skip_action = True
- move_time += time.time() - start_time
- self.unmute()
- except(Exception):
- self.unmute()
- self._agentCrash(agentIndex, quiet=True)
- return
- else:
- observation = agent.observationFunction(self.state.deepCopy())
- self.unmute()
- else:
- observation = self.state.deepCopy()
- # Solicit an action
- action = None
- self.mute()
- if self.catchExceptions:
- try:
- timed_func = TimeoutFunction(agent.getAction,
- int(self.rules.getMoveTimeout(agentIndex)) - int(
- move_time))
- try:
- start_time = time.time()
- if skip_action:
- raise (TimeoutFunctionException())
- action = timed_func(observation)
- except(TimeoutFunctionException):
- print("Agent %d timed out on a single move!" % agentIndex)
- self.agentTimeout = True
- self.unmute()
- self._agentCrash(agentIndex, quiet=True)
- return
- move_time += time.time() - start_time
- if move_time > self.rules.getMoveWarningTime(agentIndex):
- self.totalAgentTimeWarnings[agentIndex] += 1
- print("Agent %d took too long to make a move! This is warning %d" % (
- agentIndex, self.totalAgentTimeWarnings[agentIndex]))
- if self.totalAgentTimeWarnings[agentIndex] > self.rules.getMaxTimeWarnings(
- agentIndex):
- print("Agent %d exceeded the maximum number of warnings: %d" % (
- agentIndex, self.totalAgentTimeWarnings[agentIndex]))
- self.agentTimeout = True
- self.unmute()
- self._agentCrash(agentIndex, quiet=True)
- self.totalAgentTimes[agentIndex] += move_time
- # print("Agent: %d, time: %f, total: %f" % (agentIndex, move_time, self.totalAgentTimes[agentIndex]))
- if self.totalAgentTimes[agentIndex] > self.rules.getMaxTotalTime(agentIndex):
- print("Agent %d ran out of time! (time: %1.2f)" % (
- agentIndex, self.totalAgentTimes[agentIndex]))
- self.agentTimeout = True
- self.unmute()
- self._agentCrash(agentIndex, quiet=True)
- return
- self.unmute()
- except(Exception):
- self.unmute()
- self._agentCrash(agentIndex)
- return
- else:
- action = agent.getAction(observation)
- self.unmute()
- # Execute the action
- self.moveHistory.append((agentIndex, action))
- if self.catchExceptions:
- try:
- self.state = self.state.generateSuccessor(agentIndex, action)
- except(Exception):
- self._agentCrash(agentIndex)
- return
- else:
- self.state = self.state.generateSuccessor(agentIndex, action)
- # Change the display
- self.display.update(self.state.data)
- ###idx = agentIndex - agentIndex % 2 + 1
- ###self.display.update( self.state.makeObservation(idx).data )
- # Allow for game specific conditions (winning, losing, etc.)
- self.rules.process(self.state, self)
- # Track progress
- if agentIndex == numAgents + 1: self.numMoves += 1
- # Next agent
- agentIndex = (agentIndex + 1) % numAgents
- if _BOINC_ENABLED:
- boinc.set_fraction_done(self.getProgress())
- # inform a learning agent of the game result
- for agent in self.agents:
- if "final" in dir(agent):
- try:
- self.mute()
- agent.final(self.state)
- self.unmute()
- except(Exception):
- if not self.catchExceptions: raise
- self.unmute()
- print("Exception", data)
- self._agentCrash(agent.index)
- return
- self.display.finish()
|