Current Research
"To follow knowledge like a sinking star,
Beyond the utmost bound of human thought." - Ulysses by Alfred Tennyson
Home
Current Research
Previous Research
Publications
Patents
Courses
Resume
e-mail comments/questions to:
smm AT eecs DOT berkeley DOT edu
Historically, the Federal Communication Commission (FCC) has focused on the command and control regime for regulating spectrum in which the FCC allocates frequencies for specific use with power limitations, service restrictions and build-out requirements. This regulatory regime has led to an dichotomy where on one hand spectrum is a scarce resource with the same frequency being allocated multiple times, while on the other hand actual measurements show sub 1% utilization.
One of the main reasons for the disconnect between allocation and use is that the spatial and temporal scales of regulation do not match the spatial and temporal scale of actual use. The spatial scale of the current regulation regime is countries or continents while its temporal scale is years or decades. Use on the other hand, occurs spatially on the scale of coffee houses and cities and temporally on the scale of seconds/milli seconds.
One of the ways to bridge this gap is opportunistic spectrum sharing. Under such a regime, secondary users are allowed to operate in frequency bands without the consent of the primary users (users that have been allocated the spectrum by the FCC) of these bands, as long as they do not interfere with the primary user. The FCC has already legalized this type of sharing in the 5.4GHz band (5.470GHz to 5.725GHz). Devices in this spectrum range have to periodically sense for the presence of military radars. Finally, on November 4th 2008, the FCC decided to allowed secondary devices (called white space devices) to operate in the TV bands.
My research is focused on identifying key issues in the design of opportunistic radios systems. I have tried to gain an understanding of these issues using mathematical analysis/simulations and by building actual systems and experimental testbeds. In the next few sections I will explain the various sub themes of my research:
Cooperative Spectrum Sensing: I have looked at cooperative spectrum sensing as a way of reducing sensitivity requirements for individual radios. I also investigated the impact of channel correlation and misbehaving radios on cooperative gains. Details of this work can be found here. The choice of various data fusion schemes on cooperative gains was investigated here. The Achilles heel of cooperative is uncertainty in channel correlation. This impact of uncertainty in channel correlation was studied here and here.
Metrics for white space recovery: White space usage is essentially an area recovery problem. The traditional detection metrics of probability of detection (PD), probability of false alarm (PFA) and sensing time do not capture the impact of different algorithms on this area recovery problem. We proposed the metrics of Fear of Harmful Interference (FHI) and Weighted Probability of Area Recovered (WPAR) which take into account the skepticism of the primary users and are capable to evaluating cooperative sensing and assisted detection. Details of this work can be found here and here.
Evaluating the FCC's rules for white space usage: For business strategists and economists, the key question is "How much white space is there?". We analyzed the FCC's rules and evaluated the amount of white space resulting from the FCC's ruling. We define white space as the average number TV channels available per location/person. Based on the FCC's rules, an average of ~6 channels are available per person using geo-location techniques while sensing results in an average of ~0.25 channels per person. We also propose a principled way for regulators to choose the protection margin for primaries that can be eroded by secondary operation. This approach quantifies the political tradeoff between person-channels gained for potential whitespace usage versus person-channels lost for broadcasters as we vary the protection margin. For the choice of protection margin(s) used by the FCC (~1dB), the overall tradeoff is at least 30:1 while being approximately 3:1 at the margin — that is three additional people gain a channel for potential white-space use for every additional person that potentially loses reception of a channel of broadcast television. This analysis uses the ITU propagation models, the FCC high and low power transmitter databases and the US Census data of 2000. Details of this investigation can be found here. The code and data for this analysis is available for download. A version of this analysis using the FCC database at 08/15/08 can be found here.
Spectrum Sensing and system design: Spectrum sensing algorithms (Energy detection, Coherent detection and Cyclo-stationary detection) were investigated here. We used these techniques in a UWB chip to enable it to detect and avoid WiMax devices. The implementation aspects of these techniques and the overall system performance can be found here. A more technical description can be found here.
Assisted Detection: The use of transmitters in nearby bands can aid the detection of a given primary. Multiple transmitters reside on a single tower. Furthermore, the shadowing seen by these transmitters is not very frequency selective. Multiband sensing leverages this fact to reduce the impact of the shadowing uncertainty on the non-interference guarantee given to the primary users. Furthermore, cooperation among multiband sensors can behave qualitatively differently from cooperation among single-band sensors. While cooperation among single band sensors is needed to meet a target probability of harmful interference to primaries, cooperation for multiband sensors can be used instead to reduce the probability of missed opportunity for cognitive radios. These ideas are developed here and here.
Coexistence with primaries of different scales: The scale (power/height) of secondary transmitters influences the scale of the secondary devices that can co-exist with them. This idea in the context of wireless microphones is investigated here.
Channel allocation for cognitive radios: In this class project, I investigate the allocation of unoccupied channels to a group of secondary users. The criteria is to minimize the impact of a returning primary user. In the low SNR regime this allocation problem can be modeled as a max flow-min cut problem. Details can be found here.
Currently, I am investigating the following problems related to opportunistic spectrum sharing:
Co-design of sensing and MAC-layers for Cognitive radios: The idea is to use the high false alarm rates of current day Cognitive radios to design ALOHA and CSMA style MAC layers. This work is in collaboration with Prof. Murat Torlak.
Is GPS required for fixed white space devices?: The FCC requires a 50m location accuracy of fixed white space devices. Is this the right number? What is the appropriate tradeoff between available white space and location accuracy? Can we achieve reasonable location accuracy using other techniques?
Over the air TV experiments: We are finalizing a dual antenna sensing platform to sense for TV signals. (Picture of the setup coming soon!).