Minimizing cost of redundant sensor-systems with non-monotone and monotone search algorithms

This paper considers distributed multisensor applications, and finds redundant configurations that minimize system weight or cost while insuring system dependability. Given choices among different component types, fulfilling the system's operational requirements but having different dependability parameters and per item cost, two heuristics, one nonmonotone (tabu search) and one monotone (simulated annealing), are used to find configurations that minimize the chosen cost metric. The search is limited to a surface delimiting the solution space region fulfilling system dependability requirements. Experimental results are presented with cost savings of 20% compared with the least expensive system consisting of only one component type. A test case compares results from the two methods with an exhaustive search to verify that the heuristics provide reasonable solutions.