Ryan W. Thomas

Cognitive networks: adaptation and learning to achieve end-to-end performance objectives

In this article we advance the idea of a cognitive network, capable of perceiving current network conditions and then planning, learning, and acting according to end-to-end goals. Cognitive networks are motivated by the complexity, heterogeneity, and reliability requirements of tomorrows networks, which are increasingly expected to self-organize to meet user and application objectives. We compare and contrast cognitive networks with related research on cognitive radios and cross-layer design. By defining cognitive networks, examining their relationship to other technologies, discussing critical design issues, and providing a framework for implementation, we aim to establish a foundation for further research and discussion

Rendezvous for Cognitive Radios

Cognitive radios have been touted as a solution to communicating in a Dynamic Spectrum Access environment. This paper examines how cognitive radios initially find one another among the expanse of ever-changing open spectrum, termed the rendezvous problem. Specifically, it addresses the problem of rendezvous under varying levels of system capabilities, spectrum policies, and environmental conditions. The focus is on rendezvous when there are are no control channels or centralized controllers, which we term the blind rendezvous problem. Under these conditions, a sequence-based and modular clock blind rendezvous algorithms are proposed, and it is shown that the performance of these algorithms compares favorably to that of a random blind rendezvous algorithm. Specifically, the sequence-based algorithm provides a bounded Time To Rendezvous (TTR) and the ability to prioritize channels where rendezvous is more likely to occur; the modular clock algorithm reduces the expected TTR, requires little precoordination among radios attempting to rendezvous, and is robust to radios sensing different sets of available channels.

Joint Power and Channel Minimization in Topology Control: A Cognitive Network Approach

Wireless topology control is the process of structuring the connectivity between network nodes to achieve some network-wide goal. This paper presents a cognitive network approach to achieving the objectives of power and spectrum management. We cast the problem as a two phased non-cooperative game and use the properties of potential game theory to ensure the existence of, and convergence to, a desirable Nash equilibrium. Although this is a multi-objective optimization and the spectrum management problem is NP-hard, this selfish cognitive network constructs a topology that minimizes the maximum transmission power while simultaneously using, on average, less than 12% extra spectrum, as compared to the ideal solution.

The price of ignorance: distributed topology control in cognitive networks

In a cognitive network, autonomous and adaptive radios select their operating parameters to achieve individual and network-wide goals. The effectiveness of these adaptations depends on the amount of knowledge about the state of the network that is available to the radios. We examine the price of ignorance in topology control in a cognitive network with power- and spectral-efficiency objectives. We propose distributed algorithms that, if radios possess global knowledge, minimize both the maximum transmit power and the spectral footprint of the network. We show that while local (as opposed to global) knowledge has little effect on the maximum transmission power used by the network, it has a significant effect on the spectral performance. Furthermore, we show that due to the high cost of maintaining network knowledge for highly dynamic networks, the cost/performance tradeoff makes it advantageous for radios to operate under some degree of local knowledge, rather than global knowledge.We also propose distributed algorithms for power and frequency adaptations as radios join or leave the network, and assess how partial knowledge impacts the performance of these adaptations.

Algorithms and bounds for estimating location, directionality, and environmental parameters of primary spectrum users

Most existing work on dynamic spectrum access deals with creating a spectral and temporal map of spectrum white space, and then filling it. The spectrum can be better utilized by increasing the spatial awareness of secondary users to include knowledge of the locations of all primary and secondary users, as well as the orientations and parameters of their directional or omni-directional antennas. This paper derives a maximum likelihood (ML) algorithm, an approximate ML algorithm, and associated performance bounds for jointly estimating a transmitters position, orientation, beam width, and transmit power, as well as the environments path loss exponent, using received signal strength measurements. The methods can be used for either a primary or secondary user. Simulations are used to determine what types of sensor geometries lead to good estimates of each parameter, to evaluate the performance of the estimators, and to determine spectrum availability as a function of spatial coordinates.

Radio Tomography for Roadside Surveillance

Radio tomographic imaging (RTI) has recently been proposed for tracking object location via radio waves without requiring the objects to transmit or receive radio signals. The position is extracted by inferring which voxels are obstructing a subset of radio links in a dense wireless sensor network. This paper proposes a variety of modeling and algorithmic improvements to RTI for the scenario of roadside surveillance. These include the use of a more physically motivated weight matrix, a method for mitigating negative (aphysical) data due to noisy observations, and a method for combining frames of a moving vehicle into a single image. The proposed approaches are used to show improvement in both imaging (useful for human-in-the-loop target recognition) and automatic target recognition in a measured data set.

Modelling and analysis of radio tomography

Radio tomographic imaging (RTI) has recently been proposed for tracking object location via radio waves without requiring the objects to transmit or receive radio signals. The position is extracted by inferring which voxels are obstructing a subset of radio links in a dense wireless sensor network. This paper proposes a refined model for signal attenuation in RTI based on measured data, which is used to provide analytic support for previous qualitative observations. We also provide an analytic method for choosing the weighting of the regularization term, and investigate methods for dealing with negative observations caused by noise.

Optimization of Trust System Placement for Power Grid Security and Compartmentalization

This article proposes a robust mathematical method to strategically place trust nodes to compartmentalize a time-critical SCADA network. The trust nodes combine firewall and intrusion detection technology to provide communication network security for protection, control, and SCADA systems. The mathematical technique optimizes the placement of the trust nodes based on the timing requirements of existing systems and the number of trust nodes that are available in the system given constraints, which may arise due to budgetary limitations or the restrictions of existing utility hardware. The intent is to create a planning tool to allow utility system operators to determine the best locations to place trust nodes to increase system security given limited resources and/or hardware constraints. The operational requirements of the environment are translated into a mathematical model. Mixed integer linear programming is used to process this model in search of an optimal solution. Because the problem is provably NP-Hard, a heuristic is also given to quickly find good, but not optimal, solutions. Experiments show promise for the proposed techniques.

Modeling and Mitigating Noise and Nuisance Parameters in Received Signal Strength Positioning

Localization via received signal strength (RSS) is often employed in cases where the received signal is fairly weak, either due to distance or due to deliberate covert operation or interference avoidance. However, most research on source localization via RSS implicitly assumes that the background noise is negligible, and that parameters of the transmitter and environment are known. Many commercial chipsets provide per-frame RSS measurements obtained when demodulating the signal, which do not include background noise; however, noise can still cause signal outages. In law enforcement, surveillance, and emergency situations, RSS may be obtained more crudely, such as by energy detection, in which case the RSS will include contributions from the background noise as well. This paper proposes new probabilistic RSS models that account for background noise in both types of RSS measurements. We also derive and evaluate maximum likelihood estimators (MLEs) for these new models, as well as for differential RSS, which has hitherto not been rigorously analyzed in the literature. Several of these MLEs are extended to estimate the transmit power and/or path loss if they are unknown. The new models are justified by extensive measured data.

Subjective audio quality over a secure IEEE 802.11n network

This paper presents an empirical evaluation of audio quality generated by a G.711 codec and transmitted over IEEE 802.11n, IEEE 802.11b, and IEEE 802.11g Wireless Local Area Networks (WLANs). Audio quality decline due to additional calls or by securing the WLAN with Internet Protocol Security (IPsec) is quantified. Results suggest that audio quality over an IEEE 802.11n WLAN is not higher than over an IEEE 802.11b WLAN for up to 10 simultaneous calls. The data strongly suggest that toll quality audio (MOS ≥ 4.0) is not currently practical over IEEE 802.11 WLANs secured with WPA2, even using the G.711 codec.