Multilevel Data Integration with Application in Sensor Networks

We study the integration of multilevel data for applications in sensor networks. The data are assumed to be available on the different levels of a complicated stochastic system, i.e., the full system level and the subsystem level. By allowing different multivariate exponential family distributions on the data, our work not only relaxes the distribution assumptions of previous studies to a more general setting, but also provides an overall framework that is suitable for a wide range of applications. Using the maximum likelihood estimation (MLE) technique, theoretical statistical properties of the estimates, including convergence and asymptotic normally, are studied. The proposed method is applied to a sensor network for locating a target using multiple unmanned aerial vehicles. Numerical studies illustrate that the estimation accuracy is significantly improved by integrating data from multiple sources.