Memory-efficient multilevel physical optics algorithm for fast computation of scattering from three-dimensional complex targets

TL;DR: An improved version of the MLPO algorithm, for which the memory complexity is reduced to O N2 log N, so that memory growth in each step is prevented, and a significant amount of saving is achieved.

Abstract: Multilevel physical optics (MLPO) algorithm provides a speed-up for computing the physical-optics integral over complex bodies for a range of aspect angles and frequencies. On the other hand, when computation of the RCS pattern as a function of θ, φ, and frequency is desired, the O N3 memory complexity of the algorithm may prevent the solution of electrically large problems. In this paper, we propose an improved version of the MLPO algorithm, for which the memory complexity is reduced to O N2 log N. The algorithm is based on the aggregation of only some portion of the scattering patterns at each aggregation step. This way, memory growth in each step is prevented, and a significant amount of saving is achieved.