The objective of this research is to improve design of collaboratively discovering unused spectrum in a revolutionary wireless communication paradigm, the cognitive-radio network, in which cognitive-radio users can detect and share the unused spectrum. The proposed approach is to apply collaborative compressive sensing to increase spectrum sensing bandwidth, speed, and accuracy.
Specifically, the cognitive radios, rather than sweeping a set of channels sequentially, will sense linear combinations of the powers of multiple channels and report them to the fusion center, where the occupied channels are recovered using compressive sensing algorithms. Missing and erroneous reports can be exactly recovered by matrix completion since the matrix of all reports has a low-rank. Prior knowledge of channel gains is not required. The system computes more but senses much less and faster, which will be validated by both numerical and USRP2-based simulations.
The proposed research is potentially transformative as the novel framework and algorithms will broadly apply to signal sensing involving multiple sensors, modalities, and data sources. This research will have a broader impact on several audiences. The study of the jointly-sparse signal reconstruction will contribute to researchers working in compressive sensing and wireless networks. The hardware implementation will bring fresh ideas to the industrial community. The proposed research will be integrated into the existing combined education/research effort at the University of Houston and Rice University, improve education of under-represented minorities at the two institutions, and expose students to state-of-the-art research in wireless networks and compressive sensing through the NSF sponsored VIGRE program.
We are facing a "spectrum drought" today. Virtually all usable radio frequencies have been licensed. However, most of them are used ineffectively, not occupied by any so-called primary radios for long periods of time at most locations. Cognitive radios (CR) can detect licensed yet idle channels and utilize them for wireless communication. However, this requires fast spectrum sensing in a distributed manner. To the spectrum sensing challenge, a very promising solution is compressed sensing (CS). CS senses less and covers more. From a small number of so-called incoherent measurements, CS recovers signals with certain simple structures. The structures in spectrum sensing are spectrum sparsity and spatial sparsity of active channels, since most channels are silent at most locations. Spectrum sensing involves multiple CRs at spatially distributed locations, which must collaborate, or otherwise it will take each CR too long to scan for unoccupied spectrum. The intellectual merits of this project include (a) introduced CS methodology to wireless communication including collaborative spectrum sensing by CRs and OFDM channel estimation; (b) developed new sensing matrices for CRs and OFDM channel estimation; (c) developed a set of distributed computing methods for a set of collaborative computing tasks; (d) analyzed the performance of the proposed distributed algorithms by developing new analytic tools; (e) tested the proposed CS based methods for OFDM channel estimation in USRP2 prototype; [f] extended the transformative research results to other interdisciplinary fields such as smart grid, cloud computing, and big data. The broader impacts include (a) novel theoretical and algorithmic contributions for researchers working in compressed sensing and wireless networks; (b) a book on the applications of compressed sensing in wireless networking: (c) exposing graduate and undergraduate students to this research through seminar and research projects; (e) REU students and under representative students for the research projects.
