Non stationary stochastic models for forecasting QoS in ad hoc networks for real-time service support

In this paper, we try to reduce the degree of QoS degradation while enhancing at the same time the estimation of the quality of service of the networks. We shall present methods for forecasting resources to meet the QoS requirements for real-time service support based on autoregressive integrated moving average processes: ARIMA combined with the ad hoc routing protocol: DSR. These processes provide a range of models, stationary and non-stationary, that adequately represent many of the time QoS variations. The results obtained (in terms of throughputs and end-to-end delays) show that the combination of the DSR protocol with the time QoS forecasting, based on ARIMA processes, performs better than conventional DSR.