APPENDIX Deployment of Direction Finders 041713

APPENDIX
A: Comparison of Simulation-based and Numerical NPOC measures for 𝑁𝐿 = 3,000
B: Comparison of 𝑁𝑃𝑂𝐢(𝑑, πœ‹) and βˆ‘π‘–,π‘—βˆˆπΌ:𝑖<𝑗 𝑆𝑃𝑂𝐢(𝑑, 𝑙𝑖 , 𝑙𝑗 ) for Method 2 with three DFs
C: Comparison of 𝑁𝑃𝑂𝐢(𝑑, πœ‹) and βˆ‘π‘–,π‘—βˆˆπΌ:𝑖<𝑗 𝑆𝑃𝑂𝐢(𝑑, 𝑙𝑖 , 𝑙𝑗 ) for Method 2 with four DFs
D: Comparison of 𝑁𝑃𝑂𝐢(𝑑, πœ‹) and βˆ‘π‘–,π‘—βˆˆπΌ:𝑖<𝑗 𝑆𝑃𝑂𝐢(𝑑, 𝑙𝑖 , 𝑙𝑗 ) for Method 2 with five DFs
E: Comparison of 𝑁𝑃𝑂𝐢(𝑑, πœ‹) and βˆ‘π‘–,π‘—βˆˆπΌ:𝑖<𝑗 𝑆𝑃𝑂𝐢(𝑑, 𝑙𝑖 , 𝑙𝑗 ) for Method 3 with three DFs
F: Comparison of 𝑁𝑃𝑂𝐢(𝑑, πœ‹) and βˆ‘π‘–,π‘—βˆˆπΌ:𝑖<𝑗 𝑆𝑃𝑂𝐢(𝑑, 𝑙𝑖 , 𝑙𝑗 ) for Method 3 with four DFs
G: Comparison of 𝑁𝑃𝑂𝐢(𝑑, πœ‹) and βˆ‘π‘–,π‘—βˆˆπΌ:𝑖<𝑗 𝑆𝑃𝑂𝐢(𝑑, 𝑙𝑖 , 𝑙𝑗 ) for Method 3 with five DFs
H: Modified label-setting algorithm (LSA)
STEP0.
STEP1.
(i)
(ii)
(iii)
(iv)
Set 𝐿0 = {(0,0)} and 𝐿𝑖 = βˆ… for all 𝑖 ∈ 𝑁\{0}
Initialize 𝐸𝑖 accordingly for each 𝑖 ∈ 𝑁
Set 𝑇𝑖 = βˆ… for each 𝑖 ∈ 𝑁
for 𝑖 = 0,1, … , π‘š
while 𝐸𝑖 \𝑇𝑖 β‰  βˆ…
choose π‘˜ ∈ 𝐸𝑖 \𝑇𝑖 so that π‘Šπ‘–π‘˜ is minimal
for all (𝑖, 𝑗) ∈ 𝐴 with π‘Šπ‘–π‘˜ + 𝑀𝑖𝑗 ≀ 𝑛
if (π‘Šπ‘–π‘˜ + 𝑀𝑖𝑗 , πΆπ‘–π‘˜ + 𝑐𝑖𝑗 ) is not dominated
by (π‘Šπ‘—π‘™ , 𝐢𝑗𝑙 ) for any 𝑙 ∈ 𝐸𝑗
then set 𝐿𝑗 = 𝐿𝑗 βˆͺ {(π‘Šπ‘–π‘˜ + 𝑀𝑖𝑗 , πΆπ‘–π‘˜ + 𝑐𝑖𝑗 )}
if π‘Ύπ’Œπ’Š + π’˜π’Šπ’‹ = 𝒏 and π‘ͺπ’Œπ’Š + π’„π’Šπ’‹ < π‘ˆ (upper bound)
then 𝑼 = π‘ͺπ’Œπ’Š + π’„π’Šπ’‹
update 𝐸𝑗 accordingly
set 𝑇𝑖 = 𝑇𝑖 βˆͺ {π‘˜}.
The algorithm starts with initial label (0,0) associated with start node 0 and extends the set
of all labels by treating an existing label associated with a selected node. For each node 𝑖 ∈
𝑀\{π‘š + 1}, the algorithm treats all labels associated with that node (STEP1 (i)), selecting a
label with minimum π‘Šπ‘–π‘˜ . The treatment of label (π‘Šπ‘–π‘˜ , πΆπ‘–π‘˜ ) considers each outgoing arc (𝑖, 𝑗)
that does not exceed the number of DFs (i.e., π‘Šπ‘–π‘˜ + 𝑀𝑖𝑗 ≀ 𝑛) (STEP1 (ii)). A new label
(π‘Šπ‘–π‘˜ + 𝑀𝑖𝑗 , πΆπ‘–π‘˜ + 𝑐𝑖𝑗 ) associated with node 𝑗 is added only if it is not dominated by any other
labels on node 𝑗 (STEP1 (iii)). If the added label is weight feasible and the cumulative cost is
less than the current upper bound, the upper bound is improved (STEP 1 (iv)).
I. Results for cases with 1 × 1π‘˜π‘š2 and . 5 × .5π‘˜π‘š2 cells
(1)
(2)
(3)
(4)
(5)
METHOD 1
case
1
2
3
4
5
6
7
8
9
10
11
12
(6)
(7)
(8)
(9)
METHOD 2
(10)
(11)
(12)
METHOD 3
DFs
𝑍1βˆ—
RT (secs.)
𝑍2βˆ—
RT (secs.)
TR

𝑍3βˆ—
RT (secs.)
TR

3
3.46
1.04
3.37
0.06
16.79
2.62
3.37
0.03
34.70
2.62
4
4.19
1.78
4.19
0.18
9.88
0.00
4.19
0.03
59.27
0.00
5
4.78
1.29
4.78
0.21
6.14
0.02
4.65
0.03
43.00
2.64
3
4.41
2.89
4.36
0.15
19.25
1.22
4.36
0.02
144.35
1.22
4
5.62
8.29
5.59
0.84
9.87
0.45
5.42
0.02
414.60
3.59
5
6.49
10.71
6.47
1.78
6.02
0.21
6.19
0.02
535.70
4.63
3
3.82
1.69
3.74
0.08
21.06
1.91
3.74
0.02
84.25
1.91
4
4.59
3.49
4.55
0.35
10.06
0.89
4.54
0.02
174.55
0.90
5
5.28
2.70
5.25
0.45
6.00
0.55
5.22
0.02
135.00
1.22
3
2.83
3.11
2.81
0.15
20.73
0.71
2.81
0.03
103.67
0.71
4
3.65
9.38
3.65
0.95
9.89
0.00
3.45
0.03
312.67
5.24
5
4.22
12.29
4.13
2.07
5.94
2.06
4.06
0.03
409.77
3.86
3
2.48
2.76
2.42
0.12
22.65
2.39
2.38
0.02
138.15
3.85
4
3.24
8.26
3.21
0.81
10.18
1.16
3.09
0.02
412.95
4.77
5
3.78
10.37
3.76
1.73
5.99
0.57
3.57
0.02
518.65
5.48
3
3.22
4.09
3.14
0.20
20.14
2.43
3.12
0.03
136.30
3.12
4
4.17
12.44
4.17
1.25
9.95
0.02
4.11
0.03
414.63
1.58
5
4.92
17.17
4.90
2.87
5.98
0.47
4.73
0.03
572.30
3.87
3
14.12
562.04
13.98
28.75
19.55
0.99
13.98
0.56
1001.85
0.99
4
20.83
5105.87
19.47
523.56
9.75
6.53
19.20
0.55
9368.56
7.84
5
21.07
18673.43
20.83
3175.57
5.88
1.13
20.83
0.56
33285.97
1.10
3
17.97
574.19
17.65
29.13
19.71
1.79
17.58
0.56
1021.70
2.17
4
23.63
5480.94
23.63
554.90
9.88
0.00
22.22
0.55
10038.35
5.97
5
27.30
20303.37
27.25
3410.87
5.95
0.16
27.25
0.55
37185.66
0.16
3
16.74
589.59
16.53
30.37
19.41
1.26
16.53
0.56
1049.09
1.26
4
21.10
5347.78
20.88
546.86
9.78
1.02
20.40
0.56
9532.58
3.29
5
24.72
19346.19
24.30
3281.42
5.90
1.70
23.62
0.58
33528.93
4.44
3
11.54
775.00
11.40
39.42
19.66
1.19
11.40
0.66
1183.20
1.19
4
14.93
8227.46
14.68
840.57
9.79
1.64
14.50
0.66
12541.86
2.84
5
17.49
33938.44
16.98
5760.21
5.89
2.88
16.22
0.66
51814.42
7.24
3
10.20
532.09
9.91
27.00
19.70
2.93
9.96
0.50
1066.31
2.36
4
13.96
5237.68
13.89
532.04
9.84
0.45
13.08
0.51
10190.04
6.24
5
16.40
20491.54
16.08
3462.80
5.92
1.91
15.77
0.51
39866.81
3.83
3
14.05
614.18
13.52
31.53
19.48
3.77
12.94
0.59
1035.72
7.88
4
17.90
5680.74
17.88
582.62
9.75
0.15
17.79
0.59
9579.65
0.65
5
21.31
20344.32
20.84
3464.07
5.87
2.20
20.44
0.58
35258.78
4.08
Column 1 gives the case number. For each case, column 2 provides 𝑛, the number of DFs
to be deployed. Columns 3-4, 5-8, and 9-12 detail results associated with methods 1, 2, and 3,
respectively, giving, for each method, the objective function value (𝑍 βˆ— ), which is transformed
into the corresponding numerical POC value, and run time (RT) in seconds. In addition, for
Methods 2 and 3, columns 7 and 11 give the run time ratio (TR) relative to Method 1 and
columns 8 and 12 provide relative deviation βˆ† (in %) of objective function values in
comparison with Method 1 (column 3). For Method 3, column 10 includes times to compute
arc costs, build the model network, and run the modified LSA.
J. Comparisons of objective function values for . 5 × .5km2 cells
K. Comparisons of run times for . 5 × .5km2 cells
(a) Run times
L: Deployments of three DFs obtained from Method 3 (black dots represent DF locations selected)
#1: convex, left-skewed, 1 × 1km2
#2: convex, centered, 1 × 1km2
#3: convex, right-skewed, 1 × 1km2
#4: concave, left-skewed, 1 × 1km2
#5: concave, centered, 1 × 1km2
#6: concave, right-skewed, 1 × 1km2
#7: convex, left-skewed, . 5 × .5km2
#9: convex, right-skewed, . 5 × .5km2
#11: concave, centered, . 5 × .5km2
#8: convex, centered, . 5 × .5km2
#10: concave, left-skewed, . 5 × .5km2
#12: concave, right-skewed, . 5 × .5km2
#13: convex, left-skewed, . 25 × .25km2
#14: convex, centered, . 25 × .25km2
#15: convex, right-skewed, . 25 × .25km2
#16: concave, left-skewed, . 25 × .25km2
#17: concave, centered, . 25 × .25km2
#18: concave, right-skewed, . 25 × .25km2