A Universal Concept for Robust Solving of Shortest Path Problems in Dynamically Reconfigurable Graphs

Table 4

Results of the shortest path detection (in Figure 12) using the NAOP-simulator and Benchmarking between the NAOP paradigm and the DVHNN concept. A scenario corresponds to a specific choice of source-destination pair.

Shortest path results using NAOP-simulator

NAOP versus DVHNN

From source to destination

Sim. time (sim)

Convergence

Edges in the shortest path

Total cost of the path

NAOP (ms)

VDHNN (ms)

NAOP

DVHNN

Small weights values: the cost of an edge with index “” is “”

0.1

0.45

48.4

Yes

Yes

,

0.4

1.73

114

Yes

Yes

, , and

0.8

42.2

—

Yes

No

0.3

1

77.3

Yes

Yes

0.7

37.9

—

Yes

No

,

0.5

7.6

89.2

Yes

Yes

0.2

0.94

68.9

Yes

Yes

High weights values: the cost of an edge with index “” is “”

1000

0.54

—

Yes

No

,

4000

0.77

—

Yes

No

, , and

8000

70.3

—

Yes

No

3000

0.22

—

Yes

No

7000

85.4

—

Yes

No

,

5000

0.19

—

Yes

No

2000

0.16

—

Yes

No

In this paper, the concepts have been all implemented in Matlab on a standard PC.