George Mason University

Networks Laboratory


NSF Project (2014-2018)
"Spectrum Sensing and Resource Allocation in Cognitive Radio Networks"


Acknowledgement

This work was supported in part by the National Science Foundation under Grant No. CNS-1421689 [NeTS:Small].

NSF Project Outcomes Report Summary

Radio spectrum is fast becoming a scarce resource due to the growing proliferation of wireless devices and users. The conventional paradigm for allocating spectrum divides it into bands that are licensed to government and commercial entities. Unfortunately, such partitioning of the spectrum leads to a significant amount of wasted spectrum. Cognitive radio is an emerging technology that has the potential to reclaim unused spectrum without causing harm to legacy devices via a new paradigm that allows spectrum to be accessed and shared dynamically according to actual spectrum usage patterns.

In this research project, methodologies were developed that enable cognitive radios to efficiently and accurately detect and predict the presence of spectrum holes existing jointly in time and frequency across a wide spectrum band. The basic approach involves identifying primary user channels within the spectrum band and monitoring these channels to determine the best spectrum hole opportunities for the secondary, unlicensed users. An important feature of the approach is the application of a method called optimal computing budget allocation, borrowed from the field of simulation optimization, to monitoring the set of channels as efficiently as possible. Our approach essentially learns the characteristics of the identified primary user channels to enable accurate detection of spectrum holes. Accurate spectrum sensing is critical to avoid causing harmful interference to the primary or licensed users of the spectrum. Another aspect of the research involved the application of novel techniques for processing radio signals to improve the fidelity of spectrum sensing, particularly for weak signals in the presence of noise.

Another outcome of the project is a framework for dynamically allocating the unused spectrum harvested via spectrum sensing among a group of cognitive radios to allow them to form a communication network. The proposed spectrum sensing methodologies and spectrum allocation schemes were validated through computer simulations, as well as proof-of-concept implementations on software radio platforms. The research project advanced the field of knowledge in spectrum sensing and spectrum allocation to enable dynamic spectrum access allowing secondary users to share the radio spectrum with primary users. The results of the research will contribute to alleviating the issue of spectrum scarcity and help to usher in a new era of more dynamic sharing and usage of the radio spectrum.

As part of the educational component of the project a new graduate-level course on software-defined radio was developed, which includes a laboratory component to provide hands-on experience in programming software radios. Software-defined radios have been steadily gaining traction in the telecommunications industry in recent years. In contrast to legacy wireless devices, software radio platforms can be easily modified through software programming, making them attractive for rapid prototyping and technology transfer. A drawback of software radios is slower processing speeds, but this can be offset by offloading some of the computationally intensive processing to a hardware-programmable part of the software radio platform known as a Field Programmable Gate Array (FPGA). As their capstone senior design project, a team of undergraduate students designed and programmed algorithms on the FPGA of a software-defined radio to accelerate its performance. Software-defined radio platforms were found to be a very useful educational tool to teach wireless systems concepts to a new generation of communications engineers.


Participants/Organizations

Principal Investigator

  • Brian L. Mark, Professor of Electrical and Computer Engineering, George Mason University.

Graduate Students

  • Hanke Cheng, Ph.D. student in Electrical and Computer Engineering. Anticipated graduation in Spring 2020. Co-advised with Prof. Yariv Ephraim
  • Zheng Wang, Ph.D. student in Electrical and Computer Engineering. Anticipated graduation in Spring 2020.
  • Joseph M. Bruno, Ph.D. in Electrical and Computer Engineering, Spring 2017. Co-advised with Prof. Yariv Ephraim
  • Yuandao Sun, Ph.D. in Electrical and Computer Engineering, Fall 2015. Co-advised with Prof. Yariv Ephraim

Undergraduate Students

  • Yiwei Fang, B.S. student in Computer Engineering. Spring 2016 and Summer 2017.
  • Anatoliy Zinovyev, B.S. student in Mathematics and Computer Science. Fall 2016 and Spring 2016 URSP Scholar.
  • Qing Chen, B.S. student in Electrical Engineering. Summer 2015.
  • Yian Hu, B.S. student in Electrical Engineering. Summer 2015.

Collaborators

  • Yariv Ephraim, Professor of Electrical and Computer Engineering, George Mason University.
  • Zhi Tian, Professor of Electrical and Computer Engineering, George Mason University.
  • Chun-Hung Chen, Professor of Electrical and Computer Engineering, George Mason University.

Partner Organization

  • Michael Souryal, Communications Technology Laboratory at the National Institute of Standards and Technology (NIST).

High School Summer Intern

  • Mark Chitre, Rising senior at Chantilly High School. ASSIP Summer Intern, Summer 2018.

Goals

The project aims to develop models and algorithms for joint spectrum sensing and resource allocation in the dimensions of time, space, and frequency for cognitive radio (CR) networks. By taking into account spectrum sensing and resource allocation jointly in all three dimensions, higher spectrum utilization can be achieved with lower computational and communication overhead. The major research goals of the project are outlined as follows:
  1. Online characterization and tracking of spectrum occupancy in a multidimensional space.
  2. Efficient allocation of harvested spectrum resources to satisfy communications requirements.
  3. Efficient allocation of computational and communication resources for spectrum sensing.
  4. Implementation and validation of the proposed models and algorithms on a CR testbed.
The major educational goals of the project are as follows:
  1. Involve PhD students in research on state-of-the-art CR technologies.
  2. Involve MS and undergraduate students in CR research.
  3. Develop a new graduate-level course on software-defined radio with a laboratory component.

Activities

Research Activities

  • Joint wideband-temporal spectrum sensing: A recursive algorithm for joint wideband-temporal spectrum sensing was developed and shown to significantly outperform previous algorithms based on energy detection and edge detection. Furthermore, the algorithm effectively transforms a wideband sensing problem into a multiband sensing problem by aggregating groups of highly correlated channels to obtain a set of approximately independent channels. This work was led by a PhD student, Joe Bruno. The joint wideband-temporal spectrum sensing algorithm was implemented and tested on the CR testbed.
  • Collaborative spectrum sensing: A class of model-based collaborative spectrum sensing algorithms was developed and shown to significantly outperform previous collaborative spectrum sensing algorithms. This work was led by a PhD student, Yuandao Sun. In particular, the HBMM (hidden bivariate Markov model) soft fusion algorithm was shown to have much higher detection accuracy than the best linear soft fusion algorithm known in the literature. Moreover, the HBMM soft fusion approach was shown to have relatively good prediction accuracy, which can be leveraged to improve the performance of dynamic spectrum access.
  • Multiband sensing with noisy measurements: We developed a model for spectrum sensing of multiple channels with noisy measurements. The model for characterizing each channel is termed a Markov Modulated Gaussian Process (MMGP). We have developed this model further by allowing the underlying chain to be a bivariate Markov chain, which allows for non-exponential sojourn times of the PU channel in the active and idle states. We have been investigating the performance of spectrum sensing and prediction using the underlying bivariate Markov chain.
  • Optimal Computing Budget Allocation (OCBA) for multiband sensing: We developed an approach to multiband sensing using the OCBA methdology. By using the OCBA framework, the best channels for temporal spectrum sensing can be determined while minimizing the computational resources allocated to sensing the channels. The criterion for “best” is determined by the choice of objective function. In our initial studies, we have used the mean sojourn time of the primary user (PU) channel in the idle state as the objective function to maximize. We have been investigating alternative objective functions, such as duty cycle, in conjunction with the OCBA approach. In our prior work, the assumed PU model was a continuous-time Markov chain. We have been investigating the use of the MMGP model in conjunction with the OCBA approach to take into account noisy measurements. This requires us to consider the observed Fisher Information Matrix rather than the Fisher Information Matrix (FIM) as in our earlier work.
  • Spectrum sensing using cepstral and cyclostationary features: We investigated the use of cepstral and cyclostationary feature vectors together with HMM-based temporal spectrum sensing to improve spectrum sensing performance, particularly in channel environments with low signal-to-noise ratio (SNR). The idea is to replace the front-end of the earlier spectrum detector, which was based on scalar power measurements, by feature vector measurements based on the cepstrum. The cyclostationary spectrum has been applied to wideband spectrum sensing and modulation recognition in earlier studies, but not to HMM-based temporal spectrum sensing as proposed here. The cepstrum has been widely used in speech recognition and speech processing, but to our knowledge has not been applied to spectrum sensing. Our results show that the cepstrum-based feature detection outperforms cylostationary-based feature detection in moderate SNR scenarios, but the reverse is true in very low SNR scenarios. Cooperative relaying and resource allocation in cognitive radio networks: We investigated a new problem of resource allocation in cognitive radio networks, i.e., an extended Stackelberg game theoretic approach for downlink resource allocation in a secondary user (SU) relay communication scenario. This work extends existing work in the literature that has applied the Stackelberg to the downlink resource allocation problem with direct communications between SUs, but without relays. Our results have shown the that introduction of relays, while complicating the resource allocation problem, will result in better performance of the cognitive radio network provided the density of available SU relays sufficiently high.
  • Gaussian random field model for determining exclusion zones in CRNs: We developing a Gaussian random field model of interference over a coverage region in order to determine exclusion zones where CRs should not transmit, to avoid causing harmful interference to primary users.
  • Stackelberg game formulation of spectrum allocation in CRNs with relays: We investigated a new problem of resource allocation in cognitive radio networks, i.e., an extended Stackelberg game theoretic approach for downlink resource allocation in a secondary user (SU) relay communication scenario. This work extends existing work in the literature that has applied the Stackelberg to the downlink resource allocation problem with direct communications between SUs, but without relays Our results have shown the that introduction of relays, while complicating the resource allocation problem, will result in better performance of the cognitive radio network provided the density of available SU relays sufficiently high.
  • DARPA Spectum Collaboration Challenge (SC2): Together with other colleagues at George Mason University (Bernd-Peter Paris, Cameron Nowzari, Jill Nelson, and Song Min Kim), we participated in the DARPA Spectrum Collaboration Challenge (SC2) as an Open Track team with the team name SpectrumMason. The team successfully completed the DARPA hurdles for Open Track Teams in Dec. 2016, and received some additional funding to support students on the project from NSF in the form of a $100K EAGER grant. The DARPA SC2 project is closely related to the NeTS project, though the SC2 project is broader in scope, with more of a focus on implementation. To replicate the DARPA SC2 testbed, known as the Colosseum, locally at George Mason University, we acquired two Dell PowerEdge R730 servers and two NI/Ettus X310 servers.

Educational Activities

  • Graduate student research: Three PhD students were directly involved in various aspects of the project throughout the reporting period. One of the PhD students (Joe Bruno) worked on developing a multiband spectrum sensing approach based on the MMGP model and the OCBA methodology. Another student (Hanke Cheng) worked on temporal spectrum sensing using cyclostationary spectrum and cepstral feature vectors to HMM-based temporal spectrum sensing. A third student (Zheng Wang) work on the development of a Stackelberg game formulation of resource downlink allocation with relays in a cognitive radio network.
  • Graduate student teaching: A new graduate course on software-defined radio was developed and taught for the first time in Spring 2015 as ECE 699: Software-Defined Radio. The preparation for the course resulted in a set of lecture notes, prepared by Dr. Mark, and a GNU radio laboratory exercise manual, prepared by PhD student Joe Bruno. The course was taught for the second time in Spring 2018. It has been approved as a regular course in the ECE curriculum as ECE 631: Software-Defined Radio and will be taught next in Fall 2019.
  • Graduate student training: PhD students Hanke Cheng and Zheng Wang also participated in the DARPA SC2 effort and developed valuable skills in software-radio implementation. Hanke was involved in the implementation of the PHY and MAC layer protocols, while Zheng was involved in implementation of the collaboration protocol and network layer protocols.
  • Undergraduate student training: During summer 2015, two undergraduate ECE students, Qing Chen and Yian Hu, participated in the project by working on the GNU radio spectrum sensor code provided by NIST. In Spring 2015 and Fall 2015, four undergraduate students (Daniel Barcklow, Christopher Fortman, Richard Haeussler, Joshua Herr), as their captstone ECE Senior Design project, designed and implementing a spectrum analyzer using the GNU radio platform on a ZedBoard platform, which is based on the Xilinx Zynq System-On-Chip, and an AD9364 RF transceiver. The team's senior design project was awarded the Best Senior Design project of Fall 2015 within the Dept. of Electrical and Computer Engineering. In Summer 2016, an undergraduate student, Yiwei Fang, participated in software radio development related to the DARPA SC2 effort, focusing on the implementation of the collaboration protocol and network layer protocols, in particular, OLSR-based routing using the Quagga software router. For their capstone ECE Senior Design Project, a team of six undergraduate students (Joe Coffin, Pedro De Jesus, Thai Huynh, Alex Maxseiner, Lance Strain, Nguyen Vo) has been working, since Fall 2018, on programming the FPGA of the X310 SDR to perform hardware-accelerated physical layer functions such as frame synchronization.
  • High school student training: Mark Chitre, a student at Chantilly High School, was involved in the implementation of spectrum sensing in the software-radio platform for the SC2 effort as a Mason ASSIP (Aspiring Scientist Summer Internship Program) in Summer 2018.

Products

Ph.D. Dissertations

Books Chapter

  • J.M. Bruno, Y. Ephraim, B.L. Mark, and Z. Tian, "Spectrum Sensing Using Hidden Markov Models," in Handbook of Cognitive Radio, Ed. Wei Zhang, pp. 1-30, Springer Link, May 2017. ISBN 978-981-10-1389-8. DOI: 10.1007/978-981-10-1389-8_2-1 [Link to book chapter on Springer Link]

Journal Papers

  • J.M. Bruno and B.L. Mark, "A recursive algorithm for wideband temporal spectrum sensing," IEEE Trans. on Communications, vol. 66, no. 1, pp. 26-38, Jan. 2018. DOI: 10.1109/TCOMM.2017.2749578. [PDF]
  • Y. Sun, B.L. Mark, and Y. Ephraim, "Collaborative spectrum sensing via online estimation of hidden bivariate Markov models," IEEE Trans. on Wireless Communications, vol. 15, no. 8, pp. 5430-5439, August 2016. DOI: 10.1109/TWC.2016.2558506. [PDF]
  • Y. Sun, B.L. Mark, and Y. Ephraim, "Online parameter estimation for temporal spectrum sensing," IEEE Trans. on Wireless Communications, vol. 14, no. 8, pp. 4105-4114, Aug. 2015. DOI: 10.1109/TWC.2015.2416720. [PDF]

Conference Papers

  • H. Cheng, B.L. Mark, and Y. Ephraim, "Wideband Temporal Spectrum Sensing Using Cepstral Features," IEEE International Symposium on a World of Wireless, Mobile and Multimedia Networks (IEEE WoWMoM 2019), Washington DC, June 2019. [PDF]
  • Z. Wang and B.L. Mark, "Gaussian random field approximation for exclusion zones in cognitive radio networks," IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC'17), Montreal, Canada Oct. 2017. [PDF]
  • R.J. Cooper, B.L. Mark, D.W. Prescott, and K.L. Sauer, "Improving the design of atomic magnetometer arrays for RF interference mitigation in NQR detection of explosives," Proc. of SPIE Defense + Commercial Sensing, April 2017. DOI: 10.1117/12.2262392. [PDF]
  • J.M. Bruno, B.L. Mark, Y. Ephraim, and C.-H. Chen, "A computing budget allocation approach to multiband spectrum sensing," IEEE Wireless Communications and Networking Conference (WCNC), San Francisco, CA, Mar. 2017. [PDF]
  • J.M. Bruno, B.L. Mark, and Z. Tian, "An edge detection approach to wideband temporal spectrum sensing," Proc. of IEEE Globecom, Cognitive Radio and Networks Symposium, Washington DC, Dec. 2016. [PDF]
  • Y. Sun, B.L. Mark, and Y. Ephraim, "Collaborative spectrum sensing based on hidden bivariate Markov models," Proc. IEEE Globecom Workshop on Emerging Technologies for 5G Wireless Cellular Networks, San Diego, CA, Dec. 2015. [PDF]
  • J.M. Bruno and B.L. Mark, "A recursive algorithm for joint time-frequency wideband spectrum sensing," Proc. Int. Workshop on Smart Spectrum (IWSS), New Orleans, LA, Mar. 2015. [PDF]

Last updated: March 29, 2019.