Channel gain prediction in wireless networks based on spatial-temporal correlation

Due to the popularity of GPS-enabled Smartphones the location of mobile terminals has become widely available [1]. Aided by such location information, we propose a general model to predict the channel gain of mobile users for multiple time steps in advance. Our model exploits spatial and temporal correlation in a Bayesian framework. The framework is composed of an autoregressive process and, according to our analysis of Rayleigh fading, of a multivariate Gaussian process. Numerical results shows that the proposed algorithm (i) achieves much higher accuracy than autoregression, especially for long-term prediction, and (ii) is substantially more robust than support vector machines against localization errors.