A novel passive bandwidth reservation algorithm based on neural networks path prediction in wireless environments

In this paper, providing Quality of Service (QoS) to mobile hosts belonging to a real-time service class is the main issue and a new path prediction scheme based on Neural Networks (NNs) is proposed, in order to obtain a good management of system resources and service continuity, by making advanced bandwidth reservation on the Access Points (APs) where a generic user will move into. The prediction algorithm is based on two types of neural networks: the first one predicts the next cell by considering the current position and the direction of the host when the service request starts and the second one is recursively applied on previous predictions to obtain the sequence of predicted cells based on the past history. The choice of a "next cell" during the prediction phase is made by considering the probability that the user moves toward a specific adjacent cell. We hypothesize that users move according to the Smooth Random Mobility Model (SRMM) and the performance of the prediction system are analyzed in terms of prediction error for different hand-off events.