Application of Artificial Neural Networks on Vertical Traffic Simulation

Artificial neural networks (ANN) are a powerful tool for parallel processing of information, based on the behaviour of the human brain. They are able of learning and generalizing from instances and their ability as universal function approximators is widely known. They are especially useful in those problems without a mathematical model or in those with a complex or very specific model. In vertical traffic problems [1], there are a lot of outstanding variables as the hour capacity, the number of floors of the building, the number of inhabitants, the kind of building, the technical characteristics of the elevators (nominal speed, load capacity,…), the kind of operation (universal, duplex,…). All those inputs result in a very wide range of solutions that difficult the analysis problem. In this paper, an approach to the simulation of vertical traffic using a backpropagation neural network is described. We try to estimate the mean time a user has to wait depending on the values of some of these variables. The required data are obtained from a simulation program developed in the University of Zaragoza [2] that also acts as a tester or the network performance.