Light Ray Optimization and Its Parameter Analysis

Abstract

Inspired by the Fermatpsilas principle, A new optimization method-light ray optimization (LRO) is proposed. It is the combination of optical phenomena and optimization search. LRO stimulates the process of light ray propagating in uneven transparent medium to search the optimization. The searching trajectory is imaged as a beam of light ray in feasible region. The algorithm updates the searching direction by refraction or reflection of light ray in each iteration, and searches the new position. This paper applies the Sphere and Rosenbrock as benchmarks to analyze the influences of initialized parameters to the searching trajectory. The numerical experiments show that the light ray optimization is effective. And the appropriate initialized position and grid size can reduce the iteration number and improve the precision.