Skip to content

Latest commit

 

History

History
executable file
·
340 lines (310 loc) · 9.42 KB

Results.adoc

File metadata and controls

executable file
·
340 lines (310 loc) · 9.42 KB

VRPTW solutions

Average Results

or-tools hierarchical objective results compared to sintef

The distance results are partly better for or-tools, because these are results for the hierarchical objective prioritizing the number of vehicles. According to the hierarchical objective the sintef results are slightly better, or-tools is not always able to reduce the vehicle number as far, although a heavy vehicle-penalty was defined in the code. Nevertheless or-tools was able to produce nearly perfect results.

The or-tools results were generated using a 16 core AMD 5950x CPU, 12 optimizations performed in parallel, each optimization configured using 3 hours wall-time.

Table 1. Average distance hierarchical objective
optimizer C1 C2 R1 R2 RC1 RC2

sintef

828.4

589.9

1231.1

968.2

1406.1

1089.8

or-tools

828.5

598.7

1209.0

928.6

1385.6

1114.8

%deviation

0.02

1.5

-1.79

-4.09

-1.46

2.29

Table 2. Average number of vehicles hierarchical objective
optimizer C1 C2 R1 R2 RC1 RC2

sintef

10.0

3.0

12.18

2.88

11.57

3.33

or-tools

10.0

3.12

12.42

3.09

12.12

3.38

%deviation

0.0

4.17

1.93

7.51

4.78

1.25

or-tools single objective results compared to sintef

Configuring or-tools to optimize only for distance we see better distance values and worse vehicle numbers. The or-tools distance results are similar to the values given in http://web.cba.neu.edu/~msolomon/heuristi.htm .

Table 3. Average distance single objective
optimizer C1 C2 R1 R2 RC1 RC2

sintef

828.4

589.9

1231.1

968.2

1406.1

1089.8

or-tools

828.4

589.9

1182.6

878.0

1360.9

1005.3

%deviation

0.0

-0.0

-3.94

-9.31

-3.22

-7.75

Table 4. Average number of vehicles single objective
optimizer C1 C2 R1 R2 RC1 RC2

sintef

10.0

3.0

12.18

2.88

11.57

3.33

or-tools

10.0

3.0

13.33

5.45

13.12

6.25

%deviation

0.0

0.0

9.45

89.72

13.43

87.5

or-tools single objective results compared to galgos

Keep in mind that galgos results perform rounding of the distances, eight of these results violate our verification. Unfortunately we didn’t find other reference solutions. Nevertheless we see that the or-tools solutions are nearly perfect.

Table 5. Average distance single objective
optimizer C1 C2 R1 R2 RC1 RC2

galgos

828.4

589.9

1178.5

876.9

1338.1

1004.0

or-tools

828.4

589.9

1182.6

878.0

1360.9

1005.3

%deviation

0.0

0.0

0.35

0.13

1.7

0.13

Table 6. Average number of vehicles single objective
optimizer C1 C2 R1 R2 RC1 RC2

galgos

10.0

3.0

13.25

5.45

12.5

6.25

or-tools

10.0

3.0

13.33

5.45

13.12

6.25

%deviation

0.0

0.0

0.63

0.0

5.0

0.0

Continuous single objective results compared to or-tools

Continuous optimization is performed by a sequence of CRMF-NES and BiteOpt using together 1E7 evaluations per run, 64 runs, 32 runs performed in parallel. On an AMD 5950x 16 core CPU this takes about 7 minutes.

Compared to the or-tools result which serves as a reference we loose about 0.2% for the clustered problem instances and about 3% for the random problem instances. Contact me if you succeed in producing better results using any continuous optimizer.

Note that increasing the weight for the number of objectives doesn’t work as well as for or-tools, so we omit the hierarchical objective in this comparison.

Table 7. Average distance single objective
optimizer C1 C2 R1 R2 RC1 RC2

or-tools

828.4

589.9

1182.6

878.0

1360.9

1005.3

continuous

829.3

591.8

1221.6

909.7

1384.2

1035.2

%deviation

0.11

0.34

3.3

3.61

1.71

2.97

Table 8. Average number of vehicles single objective
optimizer C1 C2 R1 R2 RC1 RC2

or-tools

10.0

3.0

13.33

5.45

13.12

6.25

continuous

10.0

3.0

14.0

5.36

13.88

6.75

%deviation

0.0

0.0

5.0

-1.67

5.71

8.0

Continuous single objective results compared to galgos

Keep in mind that galgos results perform rounding of the distances, eight of these results violate our verification. Unfortunately we didn’t find other reference solutions. There is not much difference to the comparison with or-tools.

Table 9. Average distance single objective
optimizer C1 C2 R1 R2 RC1 RC2

galgos

828.4

589.9

1178.5

876.9

1338.1

1004.0

continuous

829.3

591.8

1221.6

909.7

1384.2

1035.2

%deviation

0.11

0.34

3.66

3.74

3.44

3.11

Table 10. Average number of vehicles single objective
optimizer C1 C2 R1 R2 RC1 RC2

galgos

10.0

3.0

13.25

5.45

12.5

6.25

continuous

10.0

3.0

14.0

5.36

13.88

6.75

%deviation

0.0

0.0

5.66

-1.67

11.0

8.0

Detailed results, hierarchical objective

Table 11. or-tools hierarchical objective results compared to sintef
problem vehicles distance % vehicles deviation % distance deviation

c101

10

828.9

0.0

0.0

c102

10

828.9

0.0

0.0

c103

10

828.1

0.0

-0.0

c104

10

825.6

0.0

0.11

c105

10

828.9

0.0

0.0

c106

10

828.9

0.0

0.0

c107

10

828.9

0.0

0.0

c108

10

828.9

0.0

0.0

c109

10

829.4

0.0

0.05

c201

3

591.6

0.0

-0.0

c202

3

591.6

0.0

-0.0

c203

3

591.2

0.0

-0.0

c204

3

593.9

0.0

0.56

c205

3

588.9

0.0

-0.0

c206

3

588.5

0.0

-0.0

c207

3

588.3

0.0

0.0

c208

4

655.9

33.33

11.48

r101

19

1651.2

0.0

0.03

r102

17

1487.0

0.0

0.06

r103

13

1303.5

0.0

0.83

r104

10

1002.1

11.11

-0.52

r105

14

1385.3

0.0

0.6

r106

12

1267.8

0.0

1.26

r107

10

1141.3

0.0

3.32

r108

10

956.9

11.11

-0.42

r109

12

1162.6

9.09

-2.69

r110

11

1106.6

10.0

-1.09

r111

11

1071.4

10.0

-2.31

r112

10

972.3

???

???

r201

4

1257.8

0.0

0.43

r202

4

1097.7

33.33

-7.89

r203

3

949.4

???

???

r204

3

753.2

50.0

-8.76

r205

3

1021.6

0.0

2.73

r206

3

916.9

0.0

1.19

r207

3

820.5

???

???

r208

2

730.5

0.0

0.51

r209

3

919.9

0.0

1.18

r210

3

956.3

0.0

1.8

r211

3

790.3

???

???

rc101

15

1632.0

7.14

-3.83

rc102

13

1528.8

8.33

-1.67

rc103

11

1326.1

0.0

5.11

rc104

10

1151.2

0.0

1.38

rc105

14

1593.0

7.69

-2.24

rc106

12

1441.5

9.09

1.18

rc107

11

1262.4

???

???

rc108

11

1149.7

10.0

0.87

rc201

4

1437.3

0.0

2.16

rc202

4

1161.3

???

???

rc203

3

1097.6

???

???

rc204

3

801.6

0.0

0.4

rc205

4

1311.9

0.0

1.1

rc206

3

1184.2

0.0

3.31

rc207

3

1085.2

0.0

2.27

rc208

3

839.1

0.0

1.33

Detailed results, single objective

Table 12. or-tools single objective results compared to sintef
problem vehicles distance % vehicles deviation % distance deviation

c101

10

828.9

0.0

0.0

c102

10

828.9

0.0

0.0

c103

10

828.1

0.0

-0.0

c104

10

824.8

0.0

0.0

c105

10

828.9

0.0

0.0

c106

10

828.9

0.0

0.0

c107

10

828.9

0.0

0.0

c108

10

828.9

0.0

0.0

c109

10

828.9

0.0

0.0

c201

3

591.6

0.0

0.0

c202

3

591.6

0.0

-0.0

c203

3

591.2

0.0

-0.0

c204

3

590.6

0.0

-0.0

c205

3

588.9

0.0

-0.0

c206

3

588.5

0.0

-0.0

c207

3

588.3

0.0

-0.0

c208

3

588.3

0.0

0.0

r101

20

1643.4

5.26

-0.45

r102

18

1472.8

5.88

-0.9

r103

14

1213.6

7.69

-6.12

r104

11

983.8

22.22

-2.34

r105

15

1360.8

7.14

-1.19

r106

13

1240.6

8.33

-0.91

r107

11

1077.5

10.0

-2.46

r108

11

953.1

22.22

-0.81

r109

13

1151.9

18.18

-3.59

r110

12

1083.5

20.0

-3.16

r111

12

1054.6

20.0

-3.84

r112

10

955.7

???

???

r201

8

1148.0

100.0

-8.34

r202

8

1036.5

166.67

-13.03

r203

6

875.6

???

???

r204

5

735.8

150.0

-10.87

r205

5

956.0

66.67

-3.86

r206

5

881.6

66.67

-2.71

r207

4

798.1

???

???

r208

4

706.2

100.0

-2.84

r209

5

859.9

66.67

-5.42

r210

6

904.8

100.0

-3.68

r211

4

755.9

???

???

rc101

17

1647.3

21.43

-2.93

rc102

14

1478.6

16.67

-4.9

rc103

12

1319.4

9.09

4.57

rc104

10

1150.9

0.0

1.35

rc105

16

1532.1

23.08

-5.97

rc106

13

1385.9

18.18

-2.73

rc107

12

1236.3

???

???

rc108

11

1137.0

10.0

-0.25

rc201

9

1265.8

125.0

-10.03

rc202

8

1096.5

???

???

rc203

5

935.2

???

???

rc204

4

786.4

33.33

-1.51

rc205

7

1157.7

75.0

-10.79

rc206

7

1054.6

133.33

-8.0

rc207

6

966.4

100.0

-8.93

rc208

4

780.1

33.33

-5.81

Table 13. Continous single objective results compared to or-tools
problem vehicles distance % vehicles deviation % distance deviation

c101

10

828.9

0.0

-0.0

c102

10

828.9

0.0

0.0

c103

10

830.2

0.0

0.26

c104

10

831.1

0.0

0.77

c105

10

828.9

0.0

-0.0

c106

10

828.9

0.0

-0.0

c107

10

828.9

0.0

-0.0

c108

10

828.9

0.0

-0.0

c109

10

828.9

0.0

-0.0

c201

3

591.6

0.0

-0.0

c202

3

591.6

0.0

0.0

c203

3

594.7

0.0

0.6

c204

3

603.0

0.0

2.09

c205

3

588.9

0.0

-0.0

c206

3

588.5

0.0

-0.0

c207

3

588.3

0.0

0.0

c208

3

588.3

0.0

0.0

r101

20

1670.4

0.0

1.64

r102

18

1501.8

0.0

1.97

r103

15

1246.5

7.14

2.71

r104

12

1024.1

9.09

4.1

r105

16

1407.9

6.67

3.46

r106

14

1289.2

7.69

3.91

r107

12

1119.3

9.09

3.88

r108

11

990.1

0.0

3.87

r109

14

1202.8

7.69

4.42

r110

13

1116.0

8.33

3.0

r111

12

1083.0

0.0

2.69

r112

11

1008.6

10.0

5.53

r201

8

1188.0

0.0

3.49

r202

6

1067.7

-25.0

3.01

r203

6

908.7

0.0

3.78

r204

5

766.7

0.0

4.2

r205

5

978.8

0.0

2.38

r206

4

918.8

-20.0

4.23

r207

4

835.5

0.0

4.69

r208

4

741.4

0.0

4.99

r209

6

883.8

20.0

2.78

r210

7

934.6

16.67

3.29

r211

4

783.1

0.0

3.59

rc101

17

1673.5

0.0

1.59

rc102

15

1490.4

7.14

0.8

rc103

13

1312.2

8.33

-0.54

rc104

11

1190.5

10.0

3.45

rc105

17

1576.5

6.25

2.9

rc106

13

1401.8

0.0

1.15

rc107

13

1258.1

8.33

1.77

rc108

12

1170.6

9.09

2.96

rc201

9

1297.6

0.0

2.51

rc202

8

1124.3

0.0

2.54

rc203

6

974.5

20.0

4.2

rc204

5

828.5

25.0

5.35

rc205

7

1176.2

0.0

1.6

rc206

7

1092.5

0.0

3.59

rc207

7

982.9

16.67

1.71

rc208

5

805.4

25.0

3.25