Michael V. Greening
Defence Res. Establishment Pacific, FMO, Victoria, BC V0S 1B0, Canada
This paper reports on the application of the optimization technique called simulated annealing to the problem of finding the best array configuration for a conventional beamformer. The question the optimization process has to answer is: ``For a given maximum array length and number of sensors, what is the placement and weighting of the elements that will give the narrowest mainlobe while keeping the sidelobes below a given level S.'' It is emphasized that the technique is fairly general, and that different optimization criteria could easily be substituted. During the search, a cost C is calculated for each randomly generated array configuration and set of weights. The cost increases with the width of the mainlobe, and is proportional to how much the sidelobes are above S. When the search is complete, the array configuration may or may not occupy the whole length that was accorded to it, and the sidelobes are all at or below S. By keeping the part of the cost function that does not change and using table look-up for the transcendental functions, the search can be made quite efficient, taking only on the order of minutes of CPU time on a VAX 9000 computer for 14 receivers.[sup a)]Present address: Datavision Computing Services Ltd., 1545 Pandora Ave., Ste. 203, Victoria, BC V8R 6R1, Canada.