A Descentralized Parallel-Distributed Observability Algorithm

FAPITALLE, F; VAZQUEZ, G.E; PONZONI, I; BRIGNOLE, N.B.

Proceedings WCCM

Springer Verlag

Año: 2002; p. 1 - 1

In this work we present a new parallel-distributed algorithm for observability analysis that is based on a descentralized Master-Supervisor-Worker architecture. From an engineering viewpoint, the final goal is to generate a specific sparse matrix pattern that gives information about the most convenient amount and location of sensors to be placed in an industrial plant. The algorithmic procedure proposed for the core module required to carry out this task basically consists in the judicious exploration of undirected graphs in order to detect subpaths that represent acceptable blocks for the desired matrix pattern. The method, which was specially designed for depth-first searches (DFS), avoids path revisiting. The routines were implemented under distributed environments using the standard Parallel Virtual Machine Library (PVM). Algorithmic performance was contrasted against a centralized parallel procedure for the same purpose. The descentralization proved to be advantageous because drastic run-time reductions were achieved.