In this paper, a new class of Gaussian processes is proposed for resource-constrained mobile sensor networks. Such a Gaus sian process builds on a GMRF with respect to a proximity graph over a surveillance region. The main advantages of using this class of Gaussian processes over standard Gaussian processes defined by mean and covariance functions are its numerical ef ficiency and scalability due to its built-in GMRF and its capa bility of representing a wide range of non-stationary physical processes. The formulas for Bayesian posterior predictive statis tics such as prediction mean and variance are derived and a se quential field prediction algorithm is provided for sequentially sampled observations. For a special case using compactly sup ported kernels, we propose a distributed algorithm to implement field prediction by correctly fusing all observations in Bayesian statistics. Simulation results illustrate the effectiveness of our approach.
展开▼