FAST METHODS FOR EVALUATING THE ELECTRIC FIELD LEVEL IN 2D-INDOOR ENVIRONMENTS

Abstract

When estimating the electric field level in an indoor environment, the usual complexity of the geometry and its large electric size make it necessary to deal with asymptotic assumptions, also known as high frequency techniques. But, even with these assumptions, the computational complexity, and the CPU-time cost, can be very high. Considering this drawback, this paper proposes the implementation of a Networks System for fast calculations of the Electric field in 2D-indoor environments. When dealing with an indoor environment, the computational complexity when using deterministic methods even if frequency techniques are considered to calculate the field level in a given area, can be extremely high. Because of it, this paper deals with the implementation of fast methods using Artificial Neural Networks, for the reduction of CPU-time and memory resources when calculating the field coverage in an indoor environment. Artificial Neural Networks (ANNs) are defined as intelligent knowledge based systems. This means that, starting from a previous knowledge or any training information, ANNs are able to solve certain complex mathematical problems. Moreover, any ANN's structures are assumed to be universal function's approximators, (as proved in (1) for Multilayer Perceptrons MLP, and in (2) for Radial Basis Function Networks RBFN). Taking into account these characteristics, it can be derived that an ANN is able to link inputs and outputs in a problem, as accurate as we want. if ANNs are Universal Approximators for known functions, why to use them in this application? One reason is that the analytical equations that describes the problems, are not known, in a general