Maximizing multi-sensor system dependability

This paper considers distributed sensor systems and finds redundant configurations which maximize dependability while insuring the system remains within cost or weight constraints. Given different sensor modules which fulfill the system's operational requirements but have different dependability and cost parameters, efficient methods are used to find maximum dependability configurations. These methods limit the search to a constrained subspace of the problem space. It is shown that this region must contain the optimal configuration. Three heuristics: genetic algorithms, simulated annealing and tabu search are used. Experimental results are presented with dependability gains of between 10 and 15%. These test cases compare results from all methods and verify that in most cases the simulated annealing heuristic provides the best solutions.