Mission-Aware Predictive Network

This paper describes a systematic approach towards incorporating prediction theory into a Mission-aware Predictive Network (MaPN) framework. Although prior examples indicate prediction has successfully improved performance, there has been limited quantitative evaluation and definition of systematic approaches to incorporate prediction as a building block of mission-aware networks. With prediction, we conjecture that sparsity of a system is tightly correlated with its predictability, and therefore advocate sparsity as a desirable system goal for building predictive networks. Specifically, we propose to use compressive sensing technology as a building block for a MaPN, due to its well-known property in sparsity exploitation. We demonstrate our approach via study of two missions: 1) Distributed MIMO communication in a multipath RF environment, and 2) Multipath channel prediction in frequency bands that are not directly observable. In the first example, we achieve an improvement of 25% in communication capacity. In the second example, we show reduction of overheads of 4.6Mb/s in a 20MHz LTE channel example. We demonstrate the proposed system via both simulation and field experiments.