Element placement in thinned arrays using genetic algorithms

Optimization techniques based on natural evolutionary processes, known as genetic algorithms, are applied to the problem of element placement in a discrete linear lattice, which is combinatorial in nature and cannot be analytically optimized. Representation of potential array occupancy patterns as binary vectors allows iterative utilization of numerical models of differential reproduction, crossover and mutation to converge on near-optimal solutions. Results are presented for the placement of 31 elements within a linear lattice of 63 locations under the design constraints of center and end position occupancy and bilateral symmetry. “Fitness” of the configuration is given in terms of mainlobe width and sidelobe levels of the corresponding response pattern. When results of multiple optimization runs are compared with expectations of random search, genetic algorithms show greater efficiency, especially when increasingly rarer solutions are sought