An application of multi-objective particle swarm optimization to reconstruction of a layered dielectric circular cylinder

There has been much interest in ways how to solve the inverse scattering problem of estimation of characteristics of an object(electric constant, size, shape, position, etc.) using electromagnetic waves. The problem, particularly for a strongly inhomogeneous object, is often reduced to minimization of a cost function which relates to difference between the observed and the calculated scattered wave when a wave illuminates the object. To search the minimum point of more than one cost functions more efficiently, the author has proposed a novel scheme which is called split PSO. It has been shown that the split PSO saves computational cost and tends to reduce the traps at false solutions. Also adding initialization procedure has made an improvement in escape from the traps which are presumed through wandering behavior of particles around false solutions. Optimization problems with more than one cost functions are common in real world where there exist trade-offs. The problems are solved to find a set of solutions which balances the different cost functions. PSO also has been modified into multi-objective PSO (MOPSO) to solve such optimization problems. In this paper we will apply MOPSO to the problem of reconstruction of a layered dielectric circular cylinder from scattered waves using multi-frequency data.