Bio-inspired forms of communication using chemical and molecular signals
In this paradigm of molecular communication, the transmitter releases chemicals or molecules, and encodes information on some property of these signals such as their timing or concentration. The signal then propagates through the medium between the transmitter and the receiver via means such as diffusion, until it arrives at the receiver where the signal is detected and the information decoded. This new multidisciplinary field can be used for in-body communication, secrecy, networking microscale and nanoscale devices, infrastructure monitoring in smart cities and industrial complexes, as well as for underwater communications. Since these systems are fundamentally different from telecommunication systems, most techniques that have been developed over the past few decades to advance radio technology cannot be applied to them directly. This research is on how to use chemical and molecular signals to form networks of different scales (from kilometers to nanometers) by both adapting methods from modern telecommunications, and developing new techniques.
Capacity of wireless channels and networks
One of the fundamental issues in the design of a wireless communication system is how fast data can be transmitted reliably. The theoretical upper bound for speed of data transmission over a channel is called the channel capacity. The capacity of time-varying channels depends on what is known about the channel at the transmitter and/or receiver. It also depends on whether rate can be adaptive or constant, as well as whether or not service outage can be tolerated. We have investigated capacity and capacity regions of time-varying multiuser channels, cellular systems, and ad-hoc networks with one or more antennas at the transmitter(s) and receiver(s).
Fundamental distortion limits of analog to digital compression
While most works in compression and communication consider the data in its digital form as their starting point, it is often the case that the data is originated from an analog system which assumes a continuum of states. As a result, the fundamental performance of communication must also consider the data in its analog form and the limitations in representing it using digital information. This limitations can be associated with both sampling and quantization errors. Although the sampling error can be eliminated by sampling at the Nyquist rate, technology limitations may preclude us from doing so and sub-Nyquist sampling techniques have to be employed. The fundamental problem in this research is to characterize the trade-off among the accuracy of data representation, the sampling rate at which the analog process is sampled, and the bitrate in the final digital representation. This characterization not only provides us with a theoretical limit on analog to digital conversion, but also shed light on the optimal compression and sampling schemes that attain this limit. A particular important result of this research shows that the two sources of distortion are not necessarily additive – a clever choice of the sampler and the compressor can align the two errors in a way that one term is completely absorbed by the other. For instance, with only a bitrate constraint, the distortion in any compressed representation is bounded from below by the distortion-rate function. Our research shows that in this case (and without any other structure), the sampling rate can be reduced below the Nyquist rate without affecting the overall distortion (see figure).
Rethinking cells in cellular system design
Next-generation cellular system will consist of an all-IP backbone network core with hierarchical macrocells, small cells, and indoor femtocells all sharing the same limited bandwidth. Optimal resource allocation, handoff, and WiFi offload is needed to achieve the Gbps data rates and near-ubiquitous coverage demanded by users. We are investigating self-organizing network techniques, timefrequency allocation, and handoffoffload mechanisms to achieve these ambitious performance goals.
We consider massive MIMO systems, i.e. systems with a large number of transmitter and/or receiver antennas, and look into the benefits achievable in the asymptotic regime of large antennas. Our work focuses on low-complexity transmission and detection of uncoded symbols reliably through the said systems. We exploit the spatial diversity and large aperture inherent in large antenna systems (compared to single antenna systems) to provide noncoherent communication schemes, i.e. schemes where the transmitter and the receiver do not know the instantaneous communication channel, but know the statistics. Based on a characterization of the large antenna behaviour of such systems, we propose encoding and decoding schemes and investigate the competitiveness of these schemes with schemes exploiting instantaneous channel knowledge. Our investigations use techniques from optimization, error exponent analysis, space-time coding, and robust design.
We also have developed novel cryptographic schemes based on large dimensional random matrices. In these schemes the complexity involved in decoding massive MIMO systems is related to the worst-case hardness of solving standard lattice problems providing low-complexity physical-layer cryptography commensurate with the most sophisticated forms of application-layer cryptography. We are fundamentally interested in ways in which properties of massive MIMO systems can naturally be leveraged to provide security, privacy or construct other cryptographic primitives.
Software-defined wireless networks
Software-defined wireless networks provide a unified control plane for wireless hardware to dynamically optimize heterogeneous hardware across multiple spectral bands. The control plane allows for dynamic configuration of radio parameters such as channel, power control, antenna parameters to minimize interference and provide the required performance to heterogenous applications. The control plane can also provide seamless handoff between different wireless networks as well as utilize multiple networks simultaneously to provide robustness as required for demanding applications such as video and real-time control.
Green wireless system design
A large chunk of the wireless system infrastructure in place today was not built with energy efficiency in mind. However, with ever increasing demands for wireless services, this is a significant bottleneck to future expansion. We take a cross-layer approach to studying energy efficiency in the entire communication chain. We investigate the fundamental limits of energy involved in transporting bits from one point to another. We develop models of bit transport interactions across the entire communication system chain, from the circuits of the baseband processing blocks to the networks of communicating agents that they support. This allows us to investigate the tradeoffs among QoS, power consumption and cooperation requirements on wireless networks at different layers of the communication chain.
Cognitive radio networks
Spectrum is a very scarce resource in commercial deployments of wireless networks. This has led to interest in intelligent or cognitive radios which would be able to make opportunistic use of spectrum not licensed to it, thereby coexisting with legacy systems. Our research focuses on developing enabling schemes for this technology, right from coding techniques optimal for different regimes of operation to techniques which allow smarter interference management under very general conditions.
Compressed sensing in wireless systems
While the Shannon-Nyquist sampling theorem has made possible the representation of real world signals in digital domain, and is one of the most prominent theoretical foundations of the digital age, the frequency domain restriction on signals is often too restrictive. We investigate the limits of communication system design outside this paradigm and look at exploiting other prior knowledge of real world systems and channels to come up with design insights for the practicing engineer.
Heterogeneous parallel wireless networks
Wireless networks today are becoming heterogeneous by integrating multiple radios with different capabilities, protocol stacks, and spectrum allocations. The flexibility of these different radios allows for more general dynamic resource allocation, better coverage, and higher capacity. In heterogeneous networks, users can be connected to transmitters via parallel channels operating over different networks, such as a cellular network and a WiFi network. How to optimally communicate over such parallel channels is a fundamental question that we seek to address. Building on the information theory and coding techniques, our research aims to design the optimal communication scheme to meet the fundamental limit of such parallel channel, i.e., to achieve the maximum communication rate transmitted reliably (called capacity). Another fundamental question we seek to address is how to optimally utilize channel state information (CSI) feedback in such heterogeneous parallel channels. CSI feedback is very important in the multiuser communication, and the capacity gap between the case with instantaneous CSI feedback and the case with delayed CSI feedback is often very significant. A new insight on such heterogeneous parallel channels is that, delayed CSI feedback can achieve the same performance as instantaneous CSI feedback, in terms of sum degrees of freedom (prelog factor of capacity). Such insight can allow us to design a communication system to achieve the maximum communication rate with less feedback overhead.
Distributed sensing, communications, and control
This area of research focuses on the interplay between control and communication, optimal sparse sensing and state estimation for high dimensional control systems, as well as sensor placement, outage detection and cyber physical security. We looked at the problem of estimating high-dimensional autoregressive time series from measurements arising from a large number of distributed sensors. Communication or computational constraints may lead to collection of sampled noisy measurements; we investigated estimation performance with this corruption and also observed how placing structural constraints on the time series improves performance. Future directions would include performing this estimation in a distributed setting or controlling and simultaneously learning high-dimensional dynamical systems.
Reliability and security of modern power grid
Economic and social loss due to power grid outages and blackouts have been huge due to the lack of effective sensing, communications, data processing and control over the grid. As the grid becomes increasingly complex and dependent on modern information technology, its security against cyber-physical attacks is another critical issue. We provide solutions to these critical issues by leveraging the advancement in sensing devices and developing efficient and real-time power network monitoring, analysis and control systems.
Reliable and efficient integration of renewable energies
Renewable energies, notably wind and solar, play crucial roles in achieving sustainable energy supplies for our society. While renewable energies are clean and do not rely on fossil fuels, they are also variable and uncertain in nature. Such variability and uncertainty pose great challenges on energy system operation with deep renewable energy penetration. We provide a variety of solutions for reliable and efficient integration of renewable energies. Exploiting statistical diversity among difference sources of renewables, efficient market mechanisms are developed for aggregation of renewable energies, and significant reduction of their variability and uncertainty is achieved. We also evaluate the benefit from energy storage in risk reduction of renewable energies.
Communications and signal processing in bioengineering and neuroscience
We develop signal processing algorithms for neuroscience, with a focus on epilepsy, one of the most common neurological disorders. While in most cases treating epilepsy with antiepileptic drugs is successful, this is not effective for about a third of the patients. The main treatment for such patients is a surgical procedure for removal of the seizure onset zone (SOZ), the area in the brain from which the seizures originate. The main tool for accurately identifying the SOZ is electrocorticography (ECoG) recordings taken from grids or strips of electrodes placed on the cortex. In this research we design algorithms for SOZ localization, detection and prediction of epileptic seizures, using ECoG recordings. We use tools from communication theory, signal processing and statistical signal processing, control theory, information theory, machine learning, graphs theory, and time-series analysis.
We have also applied signal processing methods to separate gene expression microarray mixed cell tissue samples. For example, tumor biopsy tissues that are profiled using microarrays contain a mixture of tumor cells, infiltrating immune cells and additional microenvironmental cells such as fibroblasts. Previous studies have shown that microarray measurements are well modeled as a linear mixture of multiple signatures from different cell types. This problem is almost identical to signal processing of hyper-spectral imaging. Thus, by incorporating state of the art signal processing tools we successfully identify the cell types in the mixture, extract their individual signatures and relative proportions per sample (Zuckerman et al. PLoS Comput Biol. 2013;9(8):e1003189). Application of this method to the abundance of gene expression microarray samples that exist in publicly available repositories can be a powerful tool for the discovery of phenomena otherwise not detected in the mixed tissue samples of individual experiments.