-
Notifications
You must be signed in to change notification settings - Fork 13
/
ACO_JSSP_Novel_2_8J4M_04-16-2017_1800_bestWork1f - play.py
650 lines (573 loc) · 28.7 KB
/
ACO_JSSP_Novel_2_8J4M_04-16-2017_1800_bestWork1f - play.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
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
import sys #for maxInt
import random
from collections import defaultdict
import datetime #for Gantt (time formatting) #duration is in days
import plotly #for Gantt (full package)
import plotly.plotly as py #for Gantt
import plotly.figure_factory as ff #for Gantt
#plotly.tools.set_credentials_file(username='addejans', api_key='65E3LDJVN63Y0Fx0tGIQ') #for Gantt -- DeJans pw:O********1*
plotly.tools.set_credentials_file(username='tutrinhkt94', api_key='vSrhCEpX6ADg7esoVCbc') #for Gantt -- T.Tran pw:OaklandU
import copy #for saving best objects
############################## INITIALIZATION --- START
def initialization():
random.seed(1)
#**************************************************************************#
#!!!!!!!!!!!!!!!!!!!!!!!!!!! DATA INPUT --- START !!!!!!!!!!!!!!!!!!!!!!!!!!!#
#!!!!!!!!!!!!!!!!!!!!!!!!!!! DATA ENTRY --- START !!!!!!!!!!!!!!!!!!!!!!!!!!!#
global numJobs, numMachines, numNodes
numJobs = 8 #****** Note, the Gantt maker needs to be updated to display numJobs amount of colors also modify resetNodes()! (if changed)
numMachines = 4
numNodes = numJobs*numMachines + 2
parameterInitialization(200,700) #input number of ants and cycles //Needed before creating node object
#machine numbers are indexed starting at 1
Job1MachSeq = [1,2,3,4]
Job2MachSeq = [2,1,3,4]
Job3MachSeq = [1,3,4,2]
Job4MachSeq = [4,1,2,3]
Job5MachSeq = [2,1,4,3]
Job6MachSeq = [1,3,4,2]
Job7MachSeq = [1,4,2,3]
Job8MachSeq = [2,4,1,3]
global Node0, Node1, Node2, Node3, Node4, Node5, Node6, Node7, Node8, Node9, Node10, Node11, Node12, Node13, Node14, Node15, Node16, Node17, Node18, Node19, Node20, Node21, Node22, Node23, Node24, Node25, Node26, Node27, Node28, Node29, Node30, Node31
#NODE(dependents, duration, machine, nodeNum) //Duration is defined with unit as days, machine 0 is equivalent to the first machine (i.e. zero indexed)
#the node.num property is 1 larger than the object Node# variable name
Node0 = Node([],1,0,1) #J1
Node1 = Node([Node0],2,1,2)
Node2 = Node([Node0, Node1],3,2,3)
Node3 = Node([Node0, Node1, Node2],4,3,4)
Node4 = Node([],5,1,5) #J2
Node5 = Node([Node4],6,0,6)
Node6 = Node([Node4, Node5],7,2,7)
Node7 = Node([Node4, Node5, Node6],8,3,8)
Node8 = Node([],9,0,9) #J3
Node9 = Node([Node8],10,2,10)
Node10 = Node([Node8, Node9],11,3,11)
Node11 = Node([Node8, Node9, Node10],12,1,12)
Node12 = Node([],13,3,13) #J4
Node13 = Node([Node12],14,0,14)
Node14 = Node([Node12, Node13],15,1,15)
Node15 = Node([Node12, Node13, Node14],16,2,16)
Node16 = Node([],17,1,17) #J5
Node17 = Node([Node16],18,0,18)
Node18 = Node([Node16, Node17],19,3,19)
Node19 = Node([Node16, Node17, Node18],20,2,20)
Node20 = Node([],21,0,21) #J6
Node21 = Node([Node20],22,2,22)
Node22 = Node([Node20, Node21],23,3,23)
Node23 = Node([Node20, Node21, Node22],24,1,24)
Node24 = Node([],25,0,25) #J7
Node25 = Node([Node24],26,3,26)
Node26 = Node([Node24, Node25],27,1,27)
Node27 = Node([Node24, Node25, Node26],28,2,28)
Node28 = Node([],29,1,29) #J8
Node29 = Node([Node28],30,3,30)
Node30 = Node([Node28, Node29],31,0,31)
Node31 = Node([Node28, Node29, Node30],32,2,32)
#dummyNodes
global Source, Sink
Source = Node([],0,-1,0) #The source information will not change
sinkDependents = [Node3, Node7, Node11, Node15, Node19, Node23, Node27, Node31] #list of the last operation of each job; these will be dependents for the sink
Sink = Node(sinkDependents,0,-1,(numNodes-1)) #The sink information will not change
global NODESLIST
NODESLIST = [Source,Node0,Node1,Node2,Node3,Node4,Node5,Node6,Node7,Node8,Node9,Node10,Node11,Node12,Node13,Node14,Node15,Node16,Node17,Node18,Node19,Node20,Node21,Node22,Node23,Node24,Node25,Node26,Node27,Node28,Node29,Node30,Node31,Sink] #the NODESLIST should be appended appropriately in numerical order
Job1Nodes = [Node0,Node1,Node2,Node3] #Nodes should be added dependent to which job they're attached
Job2Nodes = [Node4,Node5,Node6,Node7]
Job3Nodes = [Node8,Node9,Node10,Node11]
Job4Nodes = [Node12,Node13,Node14,Node15]
Job5Nodes = [Node16,Node17,Node18,Node19]
Job6Nodes = [Node20,Node21,Node22,Node23]
Job7Nodes = [Node24,Node25,Node26,Node27]
Job8Nodes = [Node28,Node29,Node30,Node31]
Job1 = Jobs([0,1,2,3], Job1Nodes, 1) #Jobs(jobSequence, Nodes, num) //where num refers to the job number
Job2 = Jobs([4,5,6,7], Job2Nodes, 2)
Job3 = Jobs([8,9,10,11], Job3Nodes, 3)
Job4 = Jobs([12,13,14,15], Job4Nodes, 4)
Job5 = Jobs([16,17,18,19], Job5Nodes, 5)
Job6 = Jobs([20,21,22,23], Job6Nodes, 6)
Job7 = Jobs([24,25,26,27], Job7Nodes, 7)
Job8 = Jobs([28,29,30,31], Job8Nodes, 8)
global JOBSLIST
JOBSLIST = [Job1, Job2, Job3, Job4, Job5, Job6, Job7, Job8] #Append list appropriately in accordance to the jobs created.
#final note: sequential order is necessary and required for proper functionality at this stage.
#!!!!!!!!!!!!!!!!!!!!!!!!!!! DATA ENTRY --- END !!!!!!!!!!!!!!!!!!!!!!!!!!!#
#!!!!!!!!!!!!!!!!!!!!!!!!!!! DATA INPUT --- END !!!!!!!!!!!!!!!!!!!!!!!!!!!#
#**************************************************************************#
global ANTLIST
ANTLIST = []
global has_cycle
has_cycle = False
global smallestSoFarMakespan, bestSoFarAnt, smallestSoFarAntNum, bestSoFarANTLIST, bestSoFarNODESLIST, JOBSLIST2, bestSoFarSolutionGraph
bestSoFarANTLIST = []
bestSoFarNODESLIST = []
JOBSLIST2 = []
bestSoFarSolutionGraph = []
constructConjunctiveGraph()
global solutionGraphList
solutionGraphList = [[] for i in range(K) ]
generateSolutionGraphs()
global nextMachine
nextMachine = -1
global currentMachine
currentMachine = -1
global feasibleNodeLists
feasibleNodeLists = [[] for i in range(K)]
global T
T = [[0.2 for i in range(numNodes)] for j in range(numNodes)] #start with a small constant?
global H #Heuristic matrix --- will this be different for different types/"species" of ants?.. TBD ***********************************************!!!!!!!!!!!!!!!
H = [[0.5 for i in range(numNodes)] for j in range(numNodes)] #****************** NEEDS TO BE FIXED TO WORK WITH HEURISTICS
global machineList
machineList = [[] for i in range(numMachines)]
global cliquesVisited
cliquesVisited = [[] for i in range(K)]
generateAnts()
generateMachineLists()
############################## INITIALIZATION --- END
############################## CLASSES --- START
class Jobs:
def __init__(self, jobSequence, Nodes, num):
self.jobSequence = jobSequence
self.Nodes = Nodes
self.num = num
class Node:
def __init__(self, dependents, duration, machine, nodeNum):
self.duration = duration
self.dependents = [dependents for i in range(K)]
self.machine = machine
self.visited = False
self.num = nodeNum
self.startTime = 0
self.endTime = 0
self.scheduled = False
self.antsVisited = [False for i in range(K)]
self.name = 'name goes here' #fill in names via constructor
self.discovered = False
class Ant:
def __init__(self, num):
self.num = num #label each ant by a number 0 to (k-1)
self.tabu = []
self.position = -1
self.T = [[0 for i in range(numNodes)] for j in range(numNodes)] #pheromone matrix
self.pheromoneAccumulator = [[0 for i in range(numNodes)] for j in range(numNodes)] #accumulator
self.transitionRuleMatrix = [[0 for i in range(numNodes)] for j in range(numNodes)] #for equation 1 transition probability
self.makespan = 0
self.species = 'none'
self.cycleDetected = False
############################## CLASSES --- END
############################## OTHER --- START
def parameterInitialization(numAnts, numCycles):
global K, C, alpha, beta, rho
global Q
global Q1, Q2
alpha = 0.5 #influence of pheromone
beta = 1 - alpha #influence of heuristic
rho = 0.7 #evaporation constant
K = numAnts #number of ants
C = numCycles #number of cycles
Q1 = float(20) #**Programming Note: this must be a float in order for TSum to calculuate as float in calculatePheromoneAccumulation()
Q2 = float(5)
#EAS Procedure determination (below) of number of ants and fixed number of cycles
#K = int(numJobs/2)
#C = 1000
#Q = float(5) # //note: there is no Q1 and Q2 -- these are original
def generateAnts():
for i in range(K):
ANTLIST.append(Ant(i))
def generateMachineLists():
for i in range(numMachines):
for j in range(numNodes):
if NODESLIST[j].machine == i:
machineList[i].append(NODESLIST[j].num)
def generateSolutionGraphs():
for k in range(K):
constructConjunctiveGraph()
solutionGraphList[k] = conjunctiveGraph
def constructConjunctiveGraph():
global conjunctiveGraph
conjunctiveGraph = [[-1 for i in range(numNodes)] for j in range(numNodes)]
for job in JOBSLIST:
for seq1 in job.jobSequence:
for seq2 in job.jobSequence:
if seq1 <> seq2 and seq1+1 == seq2:
conjunctiveGraph[seq1+1][seq2+1] = NODESLIST[seq2+1].duration
for j in range(numJobs):
conjunctiveGraph[Source.num][JOBSLIST[j].Nodes[0].num] = JOBSLIST[j].Nodes[0].duration
for j in range(numJobs):
conjunctiveGraph[JOBSLIST[j].Nodes[numMachines-1].num][Sink.num] = 0
def chooseClique(antNum):
randomClique = random.randint(0,numMachines-1) #choose Clique by random - we then choose to travel on nodes based off of pheromone trails - this prevents local optimia to occur
while randomClique in cliquesVisited[antNum]:
randomClique = random.randint(0,numMachines-1)
cliquesVisited[antNum].append(randomClique)
return randomClique
def randomAssignment():
for i in range(K):
randNode = random.randint(1,numNodes-2)
ANTLIST[i].tabu.append(NODESLIST[randNode].num)
ANTLIST[i].position = randNode
NODESLIST[randNode].antsVisited[i] = True
def defineDecidabilityRule(): #UNUSED 04/15/2017 -- For Heuristics **
for ant in ANTLIST:
speciesType = random.randint(1,2)
if speciesType == 1:
ant.species = 'SPT' #Shortest Processing Time (SPT)
elif speciesType == 2:
ant.species = 'LPT' #Longest Processing Time (LPT)
############################## OTHER --- END 4,1,25,14,13,14,9,4
############################## SCHEDULING --- START
def schedule(ant):
scheduleNode(bestSoFarNODESLIST[numNodes-1],ant)
def scheduleNode(node,ant):
for proceedingNode in node.dependents[ant.num]:
if proceedingNode.scheduled == False:
scheduleNode(proceedingNode,ant)
positionNode(node,ant) #base case
node.scheduled = True
def positionNode(node,ant):
global longestProceedingTime
if len(node.dependents[ant.num])>0:
node.startTime = (bestSoFarNODESLIST[node.num].dependents[ant.num][0].startTime + node.dependents[ant.num][0].duration)
bestSoFarNODESLIST[node.num].startTime = (bestSoFarNODESLIST[node.num].dependents[ant.num][0].startTime + node.dependents[ant.num][0].duration)
for proceedingNode in node.dependents[ant.num]:
longestProceedingTime = (proceedingNode.startTime + proceedingNode.duration)
if longestProceedingTime > node.startTime:
node.startTime = longestProceedingTime
bestSoFarNODESLIST[node.num].startTime = longestProceedingTime
else: #node has no proceeding nodes and can be scheduled right away
node.startTime = 0
bestSoFarNODESLIST[node.num].startTime = 0
node.endTime = node.startTime + node.duration
bestSoFarNODESLIST[node.num].endTime = node.startTime + node.duration
############################## SCHEDULING --- END
############################## END OF CYCLE --- START
def calculatePheromoneAccumulation(ant,b):
if b == 0:
for i in range(numNodes):
for j in range(numNodes):
if i != j and solutionGraphList[ant.num][i][j] > 0:
ant.pheromoneAccumulator[i][j] = Q1/ant.makespan # calculate w.r.t. equation 3 or 4 *** //Python Note: in python 2.7 one of the two integers
# must be a float, we declared Q as a float() type in the parameterInitialization() method
elif b == 1:
for i in range(numNodes):
for j in range(numNodes):
if i != j and solutionGraphList[ant.num][i][j] > 0:
ant.pheromoneAccumulator[i][j] = Q2/ant.makespan
def updatePheromone(bestMakespan, bestAntNum):
TSum = 0
TOld = T
for i in range(numNodes):
for j in range(numNodes):
for ant in ANTLIST:
TSum += ant.pheromoneAccumulator[i][j]
T[i][j] = TSum + rho*TOld[i][j] #update T[i][j] pheromone matrix based on equation 2 *** ***c<1 accounted for in construction() [main] method***
#pheromoneAccumulator[i][j] = 0 <---- Necessary?
TSum = 0
for i in range(numNodes): #update for best ant
for j in range(numNodes):
if solutionGraphList[bestAntNum][i][j] > 0:
T[i][j] += float(float(solutionGraphList[bestAntNum][i][j])/float(bestMakespan))
def resetAnts():
nextMachine = -1
for k in range(K):
for i in range(numMachines):
cliquesVisited[k].pop()
constructConjunctiveGraph()
generateSolutionGraphs()
#**LEARNING REMARK: #cliquesVisited = [[] for i in range(K)] *** NOTE: In Python this does not reset the variable/ double array list
for ant in ANTLIST:
ant.tabu = []
ant.position = -1
ant.T = [[0 for i in range(numNodes)] for j in range(numNodes)] #pheromone matrix
ant.pheromoneAccumulator = [[0 for i in range(numNodes)] for j in range(numNodes)] #accumulator
ant.makespan = 0
ant.cycleDetected = False
currentMachine = -1
def resetNodes():
for k in range(K):
for node in NODESLIST:
node.visited = False
node.antsVisited[k] = False
for k in range(K):
Node0.dependents[k]=[] #**LEARNING REMARK: #We can't overwrite objects
Node1.dependents[k]=[Node0]
Node2.dependents[k]=[Node0, Node1]
Node3.dependents[k]=[Node0, Node1, Node2]
Node4.dependents[k]=[]
Node5.dependents[k]=[Node4]
Node6.dependents[k]=[Node4, Node5]
Node7.dependents[k]=[Node4, Node5, Node6]
Node8.dependents[k]=[]
Node9.dependents[k]=[Node8]
Node10.dependents[k]=[Node8, Node9]
Node11.dependents[k]=[Node8, Node9, Node10]
Node12.dependents[k]=[]
Node13.dependents[k]=[Node12]
Node14.dependents[k]=[Node12, Node13]
Node15.dependents[k]=[Node12, Node13, Node14]
Node16.dependents[k]=[]
Node17.dependents[k]=[Node16]
Node18.dependents[k]=[Node16, Node17]
Node19.dependents[k]=[Node16, Node17, Node18]
Node20.dependents[k]=[]
Node21.dependents[k]=[Node20]
Node22.dependents[k]=[Node20, Node21]
Node23.dependents[k]=[Node20, Node21, Node22]
Node24.dependents[k]=[]
Node25.dependents[k]=[Node24]
Node26.dependents[k]=[Node24, Node25]
Node27.dependents[k]=[Node24, Node25, Node26]
Node28.dependents[k]=[]
Node29.dependents[k]=[Node28]
Node30.dependents[k]=[Node28, Node29]
Node31.dependents[k]=[Node28, Node29, Node30]
#dummyNodes: (Source & Sink)
Source.dependents[k] = []
Sink.dependents[k] = [Node3, Node7, Node11, Node15, Node19, Node23, Node27, Node31]
############################## END OF CYCLE --- END
############################## EXPLORATION --- START
def nextOperation(ant, machNum, cycle):
findFeasibleNodes(ant, machNum)
calculateTransitionProbability(ant)
makeDecision(ant)
def findFeasibleNodes(ant,currentMachine):
global feasibleNodeLists
feasibleNodeLists = [[] for i in range(K)]
for node in NODESLIST:
if node.antsVisited[ant.num] == False: #04/04/17: Removed not()
if node.num in machineList[currentMachine]:
feasibleNodeLists[ant.num].append(node)
def calculateTransitionProbability(ant):
for node in feasibleNodeLists[ant.num]:
if node.num not in ant.tabu:
ant.transitionRuleMatrix[ant.position][node.num] = (((T[ant.position][node.num])**(alpha)) * ((H[ant.position][node.num])**(beta)))/sum((((T[ant.position][l.num])**(alpha)) * ((H[ant.position][l.num])**(beta))) for l in feasibleNodeLists[ant.num])
def makeDecision(ant):
probabilityList = [] #assign ranges between [0,1] and pick a rand num in interval.
for node in feasibleNodeLists[ant.num]:
probabilityList.append([ant.transitionRuleMatrix[ant.position][node.num]*100,node.num])
for i in range(len(probabilityList)-1):
probabilityList[i+1][0] += probabilityList[i][0]
randomSelection = random.randint(0,100)
selectedNode = -1
for i in range(len(probabilityList)-1):
if (probabilityList[i][0] <= randomSelection) and (randomSelection <= probabilityList[i+1][0]):
selectedNode = probabilityList[i+1][1]
break
elif randomSelection <= probabilityList[i][0]: #should this be a strict less than, "<" ?
selectedNode = probabilityList[i][1]
break
if selectedNode == -1:
selectedNode = probabilityList[0][1]
ant.position = selectedNode
############################## EXPLORATON --- END
############################## CONSTRUCTION PHASE --- START
def constructionPhase(): #The Probabilistic Construction Phrase of solutions begins by K ants
for c in range(C):
defineDecidabilityRule()
for node in NODESLIST: #reset nodes visited to false since this is a new cycle
for ant in ANTLIST:
node.antsVisited[ant.num] = False
for ant in ANTLIST:
skip_counter = 0
for i in range(numMachines):
currentMachine = chooseClique(ant.num) #at random
if c<1:
if ant.num == 0:
shuffledNodes = machineList[currentMachine] #Without shuffling, this *guarantees* no 'cycle' in the solution graph for c = 0.
skip_counter = 0
for x in machineList[currentMachine]:
if skip_counter%numJobs != 0:
oldAntPosition = ant.position
ant.tabu.append(x)
ant.position = x
NODESLIST[ant.position].antsVisited[ant.num] = True
if skip_counter%numJobs != 0:
NODESLIST[ant.position].dependents[ant.num].append(NODESLIST[oldAntPosition])
solutionGraphList[ant.num][oldAntPosition][ant.position] = NODESLIST[ant.position].duration
skip_counter += 1
else:
for j in range(len(machineList[currentMachine])):
if skip_counter%numJobs != 0:
moveFrom = ant.position
nextOperation(ant, currentMachine, c)
moveTo = ant.position
ant.tabu.append(moveTo)
NODESLIST[moveTo].visited = True
NODESLIST[moveTo].antsVisited[ant.num] = True
if skip_counter%numJobs != 0:
NODESLIST[moveTo].dependents[ant.num].append(NODESLIST[moveFrom])
solutionGraphList[ant.num][moveFrom][moveTo] = NODESLIST[moveTo].duration # set equal to the duration of the moving to node (puts weight on edge)
skip_counter += 1
for ant in ANTLIST:
undiscoverNodes()
global has_cycle
has_cycle = False
cycleDetector(ant)
ant.cycleDetected = has_cycle
if ant.cycleDetected == False:
if c == 0:
if ant.num == 0:
ant.makespan = getMakespan(ant) #longest path in solution graph --- this is equivalent to the makespan
elif ant.num != 0:
ant.makespan = sys.float_info.max
elif c != 0:
ant.makespan = getMakespan(ant) #longest path in solution graph --- this is equivalent to the makespan
calculatePheromoneAccumulation(ant,0)
elif ant.cycleDetected == True:
ant.makespan = sys.float_info.max
#print(' need to fill this out, so this way pheromones and stuff are filled properly for cycle ants, i.e. they contribute nothing.') #**************************************!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
smallestMakespan = sys.float_info.max
smallestMakespan,smallestMakespanAntNum = getSmallestMakespan()
calculatePheromoneAccumulation(ANTLIST[smallestMakespanAntNum],1) #reinforce the smallestMakespan ant
print 'The best makespan of cycle ' + str(c) + ' is: ', smallestMakespan
#print ANTLIST[smallestMakespanAntNum].makespan
print ANTLIST[smallestMakespanAntNum].cycleDetected
## for j in range(numNodes):
## print solutionGraphList[smallestMakespanAntNum][j]
if c>0:
if smallestMakespan < smallestSoFarMakespan:
bestSoFarAnt = copy.deepcopy(ANTLIST[smallestMakespanAntNum])
for i in range(numNodes):
bestSoFarNODESLIST[i] = copy.deepcopy(NODESLIST[i])
bestSoFarSolutionGraph.append(copy.deepcopy(solutionGraphList[bestSoFarAnt.num][i]))
for i in range(K):
bestSoFarANTLIST[i] = copy.deepcopy(ANTLIST[i])
for i in range(numJobs):
JOBSLIST2[i] = copy.deepcopy(JOBSLIST[i])
smallestSoFarMakespan = smallestMakespan
smallestSoFarAntNum = smallestMakespanAntNum
elif c == 0:
bestSoFarAnt = copy.deepcopy(ANTLIST[smallestMakespanAntNum])
for i in range(numNodes):
bestSoFarNODESLIST.append(copy.deepcopy(NODESLIST[i]))
bestSoFarSolutionGraph.append(copy.deepcopy(solutionGraphList[bestSoFarAnt.num][i]))
for i in range(K):
bestSoFarANTLIST.append(copy.deepcopy(ANTLIST[i]))
for i in range(numJobs):
JOBSLIST2.append(copy.deepcopy(JOBSLIST[i]))
smallestSoFarMakespan = smallestMakespan
smallestSoFarAntNum = smallestMakespanAntNum
updatePheromone(smallestMakespan, smallestMakespanAntNum)
resetNodes()
resetAnts()
print 'bestSoFarAnt.makespan: =', bestSoFarAnt.makespan
print '\n'
print 'The absolute best makespan =', bestSoFarAnt.makespan
for i in range(numNodes):
print i,':'
for j in range(len(bestSoFarNODESLIST[i].dependents[bestSoFarAnt.num])):
print bestSoFarNODESLIST[i].dependents[bestSoFarAnt.num][j].num
print '\n'
schedule(bestSoFarAnt)
for i in range(numNodes):
print i,':'
print 'num',bestSoFarNODESLIST[i].num
print 'start',bestSoFarNODESLIST[i].startTime
print 'end',bestSoFarNODESLIST[i].endTime
print '\n'
makeGanttChart(bestSoFarAnt)
############################## CONSTRUCTION PHASE --- END
############################## USED FOR DETECTING CYCLES --- START
def cycleDetector(ant):
global has_cycle
for node in NODESLIST:
undiscoverNodes() #sets all nodes to undiscovered
pcount = 0
S = [] #let S be a stack
S.append(node)
while len(S) > 0:
v = S.pop()
if v.discovered == False:
if v != node:
v.discovered = True
if v == node and pcount >=1:
has_cycle = True
return
for j in range(numNodes):
if solutionGraphList[ant.num][v.num][j] >= 0:
S.append(NODESLIST[j])
pcount += 1
def undiscoverNodes():
for node in NODESLIST:
node.discovered = False
############################## USED FOR DETECTING CYCLES --- END
############################## MAKESPAN --- START
def getSmallestMakespan():
smallestMakespan = sys.float_info.max #1.7976931348623157e+308
smallestMakespanAntNum = -1
for ant in ANTLIST:
if ant.makespan < smallestMakespan:
smallestMakespan = ant.makespan
smallestMakespanAntNum = ant.num
return smallestMakespan, smallestMakespanAntNum
def getMakespan(ant):
G = defaultdict(list)
edges = []
for i in range(numNodes):
for j in range(numNodes):
if solutionGraphList[ant.num][i][j] != -1:
edges.append([NODESLIST[i], NODESLIST[j]])
for (s,t) in edges:
G[s].append(t)
all_paths = DFS(G,Source)
max_len = 0
max_paths = []
max_makespan = 0
path_duration = 0
mkspnIndex_i = -1
for i in range(len(all_paths)):
path_duration = 0
for j in range(len(all_paths[i])):
path_duration += all_paths[i][j].duration
if path_duration > max_makespan:
max_makespan = path_duration
mkspnIndex_i = i
return max_makespan
def DFS(G,v,seen=None,path=None): #v is the starting node
if seen is None: seen = []
if path is None: path = [v]
seen.append(v)
paths = []
for t in G[v]:
if t not in seen:
t_path = path + [t]
paths.append(tuple(t_path))
paths.extend(DFS(G, t, seen[:], t_path))
return paths
############################## USED FOR GETTING MAKESPAN --- END
############################## GANTT --- START
def makeGanttChart(bestSoFarAnt):
#quit() #to not use more than 30 API calls per hour or 50 per day
df = []
for job in JOBSLIST: #I believe this could stay as JOBSLIST since we pass in only the node.num into the bestSoFarNODESLIST
for node in job.Nodes:
print 'Operation/Node num: ' + str(node.num) + ', Job num: ' + str(job.num) + ', Machine num: ' + str(node.machine)
s = str(datetime.datetime.strptime('2017-04-18 00:00:00', "%Y-%m-%d %H:%M:%S") + datetime.timedelta(days=bestSoFarNODESLIST[node.num].startTime))
d = datetime.datetime.strptime(s, "%Y-%m-%d %H:%M:%S") + datetime.timedelta(days=bestSoFarNODESLIST[node.num].duration)
df.append(dict(Task=str(node.machine), Start=str(s), Finish=str(d), Resource=str(job.num)))
print 'Start date: ' + str(s)
print 'End date: ' + str(d)
print '\n'
colors = {'1': 'rgb(255, 155, 0)',
#'1': 'rgb(0, 0, 0)', // this is black
#'2': (1, 0.9, 0.16),
'2': 'rgb(255, 0, 0)',
'3': 'rgb(0, 255, 0)',
'4': 'rgb(0, 0, 255)',
'5': 'rgb(255, 255, 0)',
'6': 'rgb(255, 0, 255)',
'7': 'rgb(0, 255, 255)',
#'8': 'rgb(255, 255, 255)', //this is white
'8': 'rgb(30, 30, 30)'}
#quit()
fig = ff.create_gantt(df, colors=colors, index_col='Resource', show_colorbar=True, group_tasks=True)
py.iplot(fig, filename='4J:3M; param(' + str(K) + ',' + str(C) + ')-makespan: ' + str(bestSoFarAnt.makespan) , world_readable=True)
print 'Schedule completed.'
############################## GANTT --- END
############################## Main() Method Below:
initialization()
constructionPhase()
print 'Schedule completed.'
'''
Idea for interface:
When an ant moves to a node, add dependency to node it's moving from if it's not already a dependent.
Give node a position propery for matrix. e.g., node.matrixPosition = [i][j] #look into paper about using modulus
'''