实际应用我从来没有碰到过数量超过100的TSP问题,因为一旦超过100,通常它会有其它的特别的约束条件,这样就不是典型的TSP问题。对付非典型的问题,我会另谋出路。 作者: wujianjack2 时间: 2015-7-29 21:26
我说的非常好的例子是tsplib95中的pr2392,这是一个有2392个城市的tsp问题,state of the art的求解log如下:
Host: Leiz-PC Current process id: 3988
Using random seed 99
Problem Name: pr2392
2392-city problem (Padberg/Rinaldi)
Problem Type: TSP
Number of Nodes: 2392
Rounded Euclidean Norm (CC_EUCLIDEAN)
Set initial upperbound to 381724 (from tour)
LP Value 1: 364586.000000 (0.59 seconds)
LP Value 2: 373801.992481 (1.36 seconds)
LP Value 3: 376863.285634 (2.29 seconds)
LP Value 4: 377455.478632 (3.92 seconds)
LP Value 5: 377642.800694 (5.23 seconds)
LP Value 6: 377768.062910 (7.10 seconds)
LP Value 7: 377917.966208 (10.05 seconds)
LP Value 8: 378016.160175 (16.10 seconds)
LP Value 9: 378032.000000 (17.14 seconds)
recomputing rownorms ...
LP Value 10: 378032.000000 (17.33 seconds)
New lower bound: 378032.000000
recomputing rownorms ...
LP Value 1: 378032.000000 (17.72 seconds)
New upperbound from x-heuristic: 378032.00
Final lower bound 378032.000000, upper bound 378032.000000
Exact lower bound: 378032.000000
DIFF: 0.000000
Final LP has 3361 rows, 5704 columns, 26466 nonzeros
Optimal Solution: 378032.00
Number of bbnodes: 1
Total Running Time: 20.23 (seconds)