<p>Autonomous vehicles require that all on-board processes be efficient in time, complexity and data storage. Infact, an ideal system employs multi-funcitonal models where ever possible. The research documented hereproposes that the Region-Feature Neural Network (RFNN) and the Hyper-Ellipsoid Clustering (HEC)Kohonen neural network (or HECNN) are viable pattern analysis and control engines that contribute to thesolution of a variety of problems. The theoretical development of the RFNN and HECNN, along with severalproof-of-concept applications are presented in detail. The RFNN is a feed-forward, back-propagation modelthat is more general than standard textbook models because it also accomodates receptive fields and weightsharing. The RFNN uses a modified version of adaptive learning rates, called "shocking" to reduce training timeand maintain stability. Small-scale benchmark problems like the XOR and XOP problems are used todemonstrate the utility of the "shocking" model. Due to its modularity, the RFNN allows the user to constructflexible, multi-layered, feed-forward architectures as well as add to and prune from an architecture even aftertraining has begun. The RFNN also permits the user to include previously learned features, called "analogies" tofurther expedite the training process on new problems or whenever new classes are added. The HECNN is aself-organizing neural network that incorporates hyperellipsoid clustering by use of the Mahalanobis distance tolearn elongated shapes and obtain a stochastic measurement of data-node association. The number of nodes canalso be regulated in a self-organizing manner by measuring how well each node models the statistical propertiesof its associated data. This measurement, called "compactness", determines where and whether to add neuralunits or prune them completely. We make several enhancements to the Kolmogorov-Smirnov compactness testto control the triggering of mitosis and/or pruning. Because fewer nodes are needed for an HECNN than for aKohonen that uses only Euclidean distance, the data size is smaller for the HEC Kohonen, even forhigh-dimensional problems. The large-scale pattern analysis problems presented here for the RFNN includesonar pattern recognition and outdoor landmark recognition. For the HECNN, we focus on sonar patternrecognition and (topographical) map building. Both the RFNN and the HECNN can be generalized to solve orcontribute to the solution of other pattern recognition problems. Both are also multifunctional in that theyaccommodate compact geometric motion planning (MP), self-referencing (SR) and tracking algorithms.Additionally, we propose the "traversability vector" (t-vector) as an efficient bridge between the HECNN andboth motion planning and self-referencing for mobile robots. As with the RFNN and HECNN, the t-vector is amodular and multi-functional tool that minimizes the computation requirements and data size as it detects pathobstructions, Euclidean optimal via points, and geometric beacons, as well as identify which geometric featuresare visible to sensors in environments that can be static or dynamic. Tracking is made possible with Julier andUhlmann's unscented filter. The unscented filter particularly compliments the HECNN in that it performslow-level (non-linear) tracking more efficiently and more accurately than its predecessor, the extended Kalmanfilter (EKF). By estimating and propagating error covariances through system transformations, the unscentedfilter eliminates the need to derive Jacobian matrices. The inclusion of stochastic information inherent to the HECmap rendered the JUKF an excellent tool for our HEC-based map building, global self-localization, motionplanning and low-level tracking. <P>

<p>Ali Soheil Sadri: Novel Adaptive Power and Rate Control in Third GenerationWideband CDMA Mobile Systems, under the direction of Dr. Winser Alexander.This Dissertation proposes novel adaptive power control and rate change schemesand investigates the performance of a Wideband Code Division Multiple Access (W-CDMA)system in conjunction with these adaptive techniques. In these schemes, thetransmit power and rate are adapted to the variations of the fading channel using adaptivethresholds based on the probability distribution function (pdf) of the predicted mobilechannel power values.We define a policy similar to the traditional power control technique withthresholds except that the thresholds are set based on several regions of operation in ourAdaptive Transmitter Power Control (TPC) and Adaptive Seamless Rate Change (SRC)schemes. These regions are defined by means of the probability distribution function(pdf) of the total average channel power. The pdf is initially constructed based on thehistory of the predicted channel power values derived from the long-range predictionalgorithm. These regions can be defined such that the system operates at a constant ratioof energy per bit over noise power.In a 1-user model with one channel path, the pdf of the channel power would bean exponential or chi-square function with 2 degrees of freedom. However, in a W-CDMAsystem, normally the rake receiver has several fingers. That is, at the receiver,the system either estimates or predicts the channel coefficients at each rake finger andperforms maximal ratio combining by multiplying each finger with its conjugate or chooses the ones with the highest energy and performs maximal ratio combining on theselected fingers.In a two-user system where the multi-access interference is modeled as theStandard Gaussian Approximation (SGA), the system performance and error probabilityof our W-CDMA system becomes similar to the one for our one-user system.Consequently, in a single user detector system, when all users adopt a similar policy fortheir adaptive power and rate control, the average total Multi-Access Interference (MAI)will be reduced. The resulting channel capacity of the system in this case will beincreased and the system may operate in a lower transmit power level.We evaluate the performance of these schemes using a detailed block diagramsimulation of a W-CDMA system. We model and simulate all major components of thesystem including an accurate model for realistic mobile channels. We present simulationresults to verify that the proposed novel schemes are superior to the traditionalapproaches for transmitter power control and rate change. Furthermore, our simulationresults show that our proposed techniques reduce the effect of Multi Access Interferencein a multi-user system.<P>

<p> The research was devoted to the theoretical investigation of lattice vibrations in low-dimensional heterostructures and bulk materials with strong polaronic coupling. The purpose of the research has been to develop the phonon theory for technologically-important materials such as nitrides of Ga and Al as well as to locate new phonon-related effects which can be utilized in artificially-created heterostructures. The electron-phonon interaction has been considered quantum mechanically.The main findings can be summarized briefly as follows: 1. Consideration of carrier-induced renormalization of acoustic phonon spectra in quantum wires revealed the possibility for the Peierls phase transition into a state with periodic lattice distortion and charge-density waves of macroscopic period in artificially-prepared structures. The phase diagram for this transition has been determined. An analytical dispersion relation for the coupled electron-phonon excitation has been derived.2. It is found that the drift of two-dimensional electrons in quantum wells can lead to efficient amplification (generation) of sub-THz coherent confined acoustic vibrations due to the Cerenkov effect when the velocity of the drifted electrons exceeds the sound velocity in the given medium. A theory has been developed to describe the confinement of acoustic modes propagating along the high-symmetry directions in cubic quantum wells.3. A theory of confinement of optical phonon modes in wurtzite quantum wells has been developed. A formalism has been derived for calculation of electron scattering rates in optically anisotropic (uniaxial) crystals and quantum wells. 4. From the comparison of the energy losses to the lattice as function of the carrier velocity obtained in frameworks of perturbative model and path-integral Thornber-Feynman approach it is found that perturbation theory can be applied for materials with intermediate polaronic coupling such a GaN and AlN. Moreover, the theoretical possibility of unique low-field runaway transport in these materials has been demonstrated. <P>

<p>RAMANATH, RAJEEV. Interpolation Methods for the Bayer Color Array (under the guidance of Dr. Wesley E. Snyder)Digital still color cameras working on single CCD-based systems have a mosaicked mask of color filters on the sensors. The Bayer array configuration for the filters is popularly used. This requires that the data be interpolated to recover all the scene information. Many existing interpolation (demosaicking) algorithms that can reconstruct the scene use modifications of the bilinear interpolation method, intro-ducing a variety of artifacts in the images. These algorithms have been investi-gated. A new method for restoring these color images using an optimization method known as Mean Field Annealing is introduced using a variety of image prior models. Their performance relative to existing demosaicking methods is included.<P>

<p>TOMBAK, Ali. Radio Frequency Applications of Barium Strontium Titanate Thin Film Tunable Capacitors. (Under the supervision of Amir S. Mortazawi).Properties of thin film barium strontium titanate (BST) based capacitors for RF and microwave components were studied. The capacitors were measured for their tunability, loss tangent, frequency dependence of dielectric permittivity, and behavior at large RF signal amplitudes. A nonlinear equivalent circuit model for tunable BST capacitors was developed. Analysis of a tunable low pass filter fabrication using BST capacitors along with its intermodulation distortion measurements was given. Several simulations for bandpass filters were performed. Furthermore, a periodically loaded coplanar waveguide phase shifter utilizing the BST capacitors was designed. <P>

<p>A global modeling concept for modeling microwave circuits isdescribed. This concept allows the modeling of electromagnetic (EM)and thermal effects to be included in the simulation of electroniccircuits, by viewing EM and thermal subsystems as subcircuits. Then,circuit analysis techniques are developed from a general statevariable reduction formulation. This general formulation, based on thestate variables of the nonlinear devices, allows the analysis of largemicrowave circuits because it reduces the size of the nonlinear systemof equations to be solved. One of the derived analysis techniques isbased on convolution and therefore provides modeling offrequency-defined network elements not present in conventional circuitsimulators. Another analysis technique based on wavelets that wouldenable the multiresolution analysis of circuits is investigated. Also,a reduced state variable formulation using conventional time marchingschemes is developed. It is shown that this can achieve more than anorder of magnitude improvement in simulation speed compared to that oftraditional circuit simulation methods. All these developments areimplemented in a circuit simulator program, called Transim. Thisprogram provides unprecedented flexibility for the addition of newdevice models or circuit analysis algorithms. Transim supports thelocal reference concept, which is fundamental to the analysis ofspatially distributed circuits and also to simultaneousthermal-electrical simulations. Transim is applied to the transientsimulation of a 47-section nonlinear transmission line consideringfrequency dependent attenuation for the first time and the transientsimulation, also for the first time, of two quasi-optical poweramplifier arrays.<P>

<p>This thesis work consists of two distinct parts: a study ofreal-time intrusion detection on network link-state routingprotocol attacks (Part I), and a study of source identification for spoofed IP packets (Part II). These two parts could be united into a common framework consisting of an intrusion detection system and an intrusion response system. However, in many ways they are distinct and self-contained. In Part I, a real-time knowledge-based network intrusiondetection model for a link-state routing protocol is presented to detect different attacks for the protocol. This model includes three layers: a to parse packets and dispatch data, an to abstract predefined real-time events for the link-state routing protocol, and an to express thereal-time behavior of the protocol engine and to detect the intrusions by pattern matching. The timed FSM named JiNao Finite State Machine (JFSM)</I> is extended from the conventional FSM with timed states, multiple timers, and time constraints on statetransitions. The JFSM is implemented as a generator which can createany FSM according to a description in a configuration file. Theresults show that this approach is very effective for real-timeintrusion detection. This approach can be extended for use in othernetwork protocol intrusion detection systems, especially for thosewith known attacks.In Part II, a security management framework, the Decentralized Source Identification System (DECIDUOUS)</I>, is presentedto identify the ``true'' sources of network-based intrusions. The premise of this approach is that if an attack packet has been correctly authenticated by a certain router, the attack packet must have been transmitted through that router.</I> It utilizes security associations to dynamically deploy secure authentication tunnels in order to further trace down the possible attackers' locations. We present the algorithms to support the tracing of multiple attacks launched from different locations, even across several administrative domains. Our results show that the DECIDUOUS system is reasonably efficient, flexible and robust. Our approach could serve as the basis for future research on different tracing strategies for differenttypes of attacks in large-scale networks.<P>

<p>In wireless communication systems, the direct signal andthe reflected signals form an interference pattern resulting in a received signal given by the sum of these components. They are distinguished by their Doppler shifts at the mobile. Since the parameters associated with these components are slowly varying, the fading coefficients can be accurately predicted far ahead. We introduced a novel algorithm for long range predictionof fading channels. This algorithm finds the linear Minimum MeanSquared Error (MMSE) estimate of the future fading coefficientsgiven a fixed number of previous observations. We show that thesuperior performance of this algorithm is due to its longer memoryspan achieved by using lower sampling rate given fixed model orderrelative to the conventional (data rate) methods of fadingprediction. This long range prediction capability for fading channels would provide enabling technology for accurate power control, reliable transmitter and/or receiver diversity, more effective adaptive modulation and coding and improvements in many other components ofwireless systems. In this thesis, we demonstrate that largeimprovements in the performance are possible for both flat andfrequency-selective rapidly varying fading channels when theproposed prediction method is used. We illustrate the performanceenhancements both at the transmitter and the receiver with boththeoretical and simulation results.<P>

<p> This dissertation reports results which pioneered the novel application of wavelets to processes and issues critical to semiconductor manufacturing. This work is especially germane given: 1) the rising costs and complexity of manufacturing; 2) the increasing deluge of data provided by process andequipment sensors that are developed, in part, to address these costs; and 3) the lack of adequate tools for handling the information overload. The sheer quantity of data is currently outstripping conventional means of storage and analysis; as a result, an increased fraction of this data must be discarded or incompletely processed. Ultimately, by developing advanced tools and methodologies to address these issues, cost and performance-driven decisions can be made in a timely and cost-effective manner. The content of this dissertation impacts these critical issues through the following specific contributions. Wavelet-based methodologies were developed for the localized modeling and compression of key process-relevant information, especially information pertaining to the detection of equipment and process faults. These methodologies significantly contribute to the following general problems: 1) identification of specific local features in potentially large nonstationary data sets; 2) compression of entire data sets consisting of disjoint smooth and nonsmooth segments; and 3) improvement in the quality of modeling for important local features carrying key information. This thesis addressed these problems using wavelet networks, as the tool, and the following novel approaches: 1) a novel objective function that incorporates both global and local error as well as model parsimony; 2) a novel adaptive network initialization scheme; 3) a novel application of the wavelet transform modulus maxima (WTMM) representation to help determine and prioritize the set of local features; 4) a new network construction procedure; and 5) a new errorspace analysis (ESA) technique, to assist in the visualization of both local and global network approximation errors during wavelet network construction. Information gathered in situ using a quadrupole mass spectrometer (QMS) sensor facilitated the development of a novel and unique approach for monitoring the selective film thickness during selective growth of silicon epitaxy. Moreover, this approach is applicable to the detection of one particularly critical process fault --- loss of selectivity. In particular, since QMS sensors are currently in widespread use in most vacuum systems, this technique represents a viable, cost-effective solution to selectivity loss detection that appears readily transferable to other process chemistries. In general, by correlating similar in situ process metrics (e.g., signal area) to ex situ process observables (e.g., thin-film thickness), contributions are made to: 1) thin-film metrology and metal oxide semiconductor field effect transistor (MOSFET) gate-stack engineering; 2) run-to-run detection of aberrant signalmodes, including process and equipment faults; and 3) decision-making using wavelet-compressed spaces. The methodologies developed in this thesis appear to be broadly applicable to otherdisciplines and fields of endeavor, including intelligent manufacturing and information technology. <P>

<p>JIANG, XIN. Two new Ka-band traveling wave power divider/combiner designs. (Under the direction of Dr. Amir Mortazawi).The purpose of this thesis is to develop a traveling wave power dividing/combining technique for a Ka-band 32-device high power amplifier system. Based on this technique, two main components of the power amplifier system have been designed. One component is a back-to-back connected 4-way waveguide-to-waveguide power divider/combiner and the other one is a back-to-back connected 8-way waveguide-to-microstrip power divider/combiner. Of the two components, the former has exhibited a 4 GHz (~12.5%) 0.5 dB bandwidth with 0.15 dB insertion loss at 32 GHz in simulation. The later utilizes a novel power coupling structure, which has the advantages of low profile, ease of fabrication and efficient heat sinking for the solid-state amplifiers. It was fabricated and experiments on this passive structure have demonstrated a minimum overall insertion loss of 1.8 dB at 32.5 GHz with a 3 dB bandwidth of 15%. The power amplifier composed of these two traveling-wave designs offers a potential broadband solution for solid-state amplifiers applied in high power millimeter-wave systems.<P>