Distributed Optimal Control of Cyber-Physical Systems: Controller Synthesis, Architecture Design and System Identification

The centralized paradigm of a single controller and a single plant upon which modern control theory is built is no longer applicable to modern cyber-physical systems of interest, such as the power-grid, software defined networks or automated highways systems, as these are all large-scale and spatially distributed. Both the scale and the distributed nature of these systems has motivated the decentralization of control schemes into local sub-controllers that measure, exchange and act on locally available subsets of the globally available system information. This decentralization of control logic leads to different decision makers acting on asymmetric information sets, introduces the need for coordination between them, and perhaps not surprisingly makes the resulting optimal control problem much harder to solve. In fact, shortly after such questions were posed, it was realized that seemingly simple decentralized optimal control problems are computationally intractable to solve, with the Wistenhausen counterexample being a famous instance of this phenomenon. Spurred on by this perhaps discouraging result, a concerted 40 year effort to identify tractable classes of distributed optimal control problems culminated in the notion of quadratic invariance, which loosely states that if sub-controllers can exchange information with each other at least as quickly as the effect of their control actions propagates through the plant, then the resulting distributed optimal control problem admits a convex formulation.

The identification of quadratic invariance as an appropriate means of "convexifying" distributed optimal control problems led to a renewed enthusiasm in the controller synthesis community, resulting in a rich set of results over the past decade. The contributions of this thesis can be seen as being a part of this broader family of results, with a particular focus on closing the gap between theory and practice by relaxing or removing assumptions made in the traditional distributed optimal control framework. Our contributions are to the foundational theory of distributed optimal control, and fall under three broad categories, namely controller synthesis, architecture design and system identification.

We begin by providing two novel controller synthesis algorithms. The first is a solution to the distributed H-infinity optimal control problem subject to delay constraints, and provides the only known exact characterization of delay-constrained distributed controllers satisfying an H-infinity norm bound. The second is an explicit dynamic programming solution to a two player LQR state-feedback problem with varying delays. Accommodating varying delays represents an important first step in combining distributed optimal control theory with the area of Networked Control Systems that considers lossy channels in the feedback loop. Our next set of results are concerned with controller architecture design. When designing controllers for large-scale systems, the architectural aspects of the controller such as the placement of actuators, sensors, and the communication links between them can no longer be taken as given -- indeed the task of designing this architecture is now as important as the design of the control laws themselves. To address this task, we formulate the Regularization for Design (RFD) framework, which is a unifying computationally tractable approach, based on the model matching framework and atomic norm regularization, for the simultaneous co-design of a structured optimal controller and the architecture needed to implement it. Our final result is a contribution to distributed system identification. Traditional system identification techniques such as subspace identification are not computationally scalable, and destroy rather than leverage any a priori information about the system's interconnection structure. We argue that in the context of system identification, an essential building block of any scalable algorithm is the ability to estimate local dynamics within a large interconnected system. To that end we propose a promising heuristic for identifying the dynamics of a subsystem that is still connected to a large system. We exploit the fact that the transfer function of the local dynamics is low-order, but full-rank, while the transfer function of the global dynamics is high-order, but low-rank, to formulate this separation task as a nuclear norm minimization problem. Finally, we conclude with a brief discussion of future research directions, with a particular emphasis on how to incorporate the results of this thesis, and those of optimal control theory in general, into a broader theory of dynamics, control and optimization in layered architectures.

Exploring the Dynamic Radio Sky: The Search for Slow Transients with the VLA

While synoptic surveys in the optical and at high energies have revealed a rich discovery phase space of slow transients, a similar yield is still awaited in the radio. Majority of the past blind surveys, carried out with radio interferometers, have suffered from a low yield of slow transients, ambiguous transient classifications, and contamination by false positives. The newly-refurbished Karl G. Jansky Array (Jansky VLA) offers wider bandwidths for accurate RFI excision as well as substantially-improved sensitivity and survey speed compared with the old VLA. The Jansky VLA thus eliminates the pitfalls of interferometric transient search by facilitating sensitive, wide-field, and near-real-time radio surveys and enabling a systematic exploration of the dynamic radio sky. This thesis aims at carrying out blind Jansky VLA surveys for characterizing the radio variable and transient sources at frequencies of a few GHz and on timescales between days and years. Through joint radio and optical surveys, the thesis addresses outstanding questions pertaining to the rates of slow radio transients (e.g. radio supernovae, tidal disruption events, binary neutron star mergers, stellar flares, etc.), the false-positive foreground relevant for the radio and optical counterpart searches of gravitational wave sources, and the beaming factor of gamma-ray bursts. The need for rapid processing of the Jansky VLA data and near-real-time radio transient search has enabled the development of state-of-the-art software infrastructure. This thesis has successfully demonstrated the Jansky VLA as a powerful transient search instrument, and it serves as a pathfinder for the transient surveys planned for the SKA-mid pathfinder facilities, viz. ASKAP, MeerKAT, and WSRT/Apertif.

New technologies driving decade-bandwidth radio astronomy : quad-ridged flared horn and compound-semiconductor LNAs

Among the branches of astronomy, radio astronomy is unique in that it spans the largest portion of the electromagnetic spectrum, e.g., from about 10 MHz to 300 GHz. On the other hand, due to scientific priorities as well as technological limitations, radio astronomy receivers have traditionally covered only about an octave bandwidth. This approach of "one specialized receiver for one primary science goal" is, however, not only becoming too expensive for next-generation radio telescopes comprising thousands of small antennas, but also is inadequate to answer some of the scientific questions of today which require simultaneous coverage of very large bandwidths.

This thesis presents significant improvements on the state of the art of two key receiver components in pursuit of decade-bandwidth radio astronomy: 1) reflector feed antennas; 2) low-noise amplifiers on compound-semiconductor technologies. The first part of this thesis introduces the quadruple-ridged flared horn, a flexible, dual linear-polarization reflector feed antenna that achieves 5:1-7:1 frequency bandwidths while maintaining near-constant beamwidth. The horn is unique in that it is the only wideband feed antenna suitable for radio astronomy that: 1) can be designed to have nominal 10 dB beamwidth between 30 and 150 degrees; 2) requires one single-ended 50 Ohm low-noise amplifier per polarization. Design, analysis, and measurements of several quad-ridged horns are presented to demonstrate its feasibility and flexibility.

The second part of the thesis focuses on modeling and measurements of discrete high-electron mobility transistors (HEMTs) and their applications in wideband, extremely low-noise amplifiers. The transistors and microwave monolithic integrated circuit low-noise amplifiers described herein have been fabricated on two state-of-the-art HEMT processes: 1) 35 nm indium phosphide; 2) 70 nm gallium arsenide. DC and microwave performance of transistors from both processes at room and cryogenic temperatures are included, as well as first-reported measurements of detailed noise characterization of the sub-micron HEMTs at both temperatures. Design and measurements of two low-noise amplifiers covering 1--20 and 8—50 GHz fabricated on both processes are also provided, which show that the 1--20 GHz amplifier improves the state of the art in cryogenic noise and bandwidth, while the 8--50 GHz amplifier achieves noise performance only slightly worse than the best published results but does so with nearly a decade bandwidth.

The relationship between the radio and gamma-ray emission of blazars

Blazars are active galaxies with a jet closely oriented to our line of sight. They are powerful, variable emitters from radio to gamma-ray wavelengths. Although the general picture of synchrotron emission at low energies and inverse Compton at high energies is well established, important aspects of blazars are not well understood. In particular, the location of the gamma-ray emission region is not clearly established, with some theories favoring a location close to the central engine, while others place it at parsec scales in the radio jet.

We developed a program to locate the gamma-ray emission site in blazars, through the study of correlated variations between their gamma-ray and radio-wave emission. Correlated variations are expected when there is a relation between emission processes at both bands, while delays tell us about the relative location of their energy generation zones. Monitoring at 15 GHz using the Owens Valley Radio Observatory 40 meter telescope started in mid-2007. The program monitors 1593 blazars twice per week, including all blazars detected by the Fermi Gamma-ray Space Telescope (Fermi) north of -20 degrees declination. This program complements the continuous monitoring of gamma-rays by Fermi.

Three year long gamma-ray light curves for bright Fermi blazars are cross-correlated with four years of radio monitoring. The significance of cross-correlation peaks is investigated using simulations that account for the uneven sampling and noise properties of the light curves, which are modeled as red-noise processes with a simple power-law power spectral density. We found that out of 86 sources with high quality data, only three show significant correlations (AO 0235+164, B2 2308+34 and PKS 1502+106). Additionally, we find a significant correlation for Mrk 421 when including the strong gamma-ray/radio flare of late 2012. In all four cases radio variations lag gamma-ray variations, suggesting that the gamma-ray emission originates upstream of the radio emission. For PKS 1502+106 we locate the gamma-ray emission site parsecs away from the central engine, thus disfavoring the model of Blandford and Levinson (1995), while other cases are inconclusive. These findings show that continuous monitoring over long time periods is required to understand the cross-correlation between gamma-ray and radio-wave variability in most blazars.

Diffraction-limited imaging on the 200-inch telescope

We have used the technique of non-redundant masking at the Palomar 200-inch telescope and radio VLBI imaging software to make optical aperture synthesis maps of two binary stars, β Corona Borealis and σ Herculis. The dynamic range of the map of β CrB, a binary star with a separation of 230 milliarcseconds is 50:1. For σ Her, we find a separation of 70 milliarcseconds and the dynamic range of our image is 30:1. These demonstrate the potential of the non-redundant masking technique for diffraction-limited imaging of astronomical objects with high dynamic range.

We find that the optimal integration time for measuring the closure phase is longer than that for measuring the fringe amplitude. There is not a close relationship between amplitude errors and phase errors, as is found in radio interferometry. Amplitude self calibration is less effective at optical wavelengths than at radio wavelengths. Primary beam sensitivity correction made in radio aperture synthesis is not necessary in optical aperture synthesis.

The effects of atmospheric disturbances on optical aperture synthesis have been studied by Monte Carlo simulations based on the Kolmogorov theory of refractive-index fluctuations. For the non-redundant masking with τ_c-sized apertures, the simulated fringe amplitude gives an upper bound of the observed fringe amplitude. A smooth transition is seen from the non-redundant masking regime to the speckle regime with increasing aperture size. The fractional reduction of the fringe amplitude according to the bandwidth is nearly independent of the aperture size. The limiting magnitude of optical aperture synthesis with τ_c-sized apertures and that with apertures larger than τ_c are derived.

Monte Carlo simulations are also made to study the sensitivity and resolution of the bispectral analysis of speckle interferometry. We present the bispectral modulation transfer function and its signal-to-noise ratio at high light levels. The results confirm the validity of the heuristic interferometric view of image-forming process in the mid-spatial-frequency range. The signal-to- noise ratio of the bispectrum at arbitrary light levels is derived in the mid-spatial-frequency range.

The non-redundant masking technique is suitable for imaging bright objects with high resolution and high dynamic range, while the faintest limit will be better pursued by speckle imaging.

Analysis-aware design of embedded systems software

In the past many different methodologies have been devised to support software development and different sets of methodologies have been developed to support the analysis of software artefacts. We have identified this mismatch as one of the causes of the poor reliability of embedded systems software. The issue with software development styles is that they are ``analysis-agnostic.'' They do not try to structure the code in a way that lends itself to analysis. The analysis is usually applied post-mortem after the software was developed and it requires a large amount of effort. The issue with software analysis methodologies is that they do not exploit available information about the system being analyzed.

In this thesis we address the above issues by developing a new methodology, called "analysis-aware" design, that links software development styles with the capabilities of analysis tools. This methodology forms the basis of a framework for interactive software development. The framework consists of an executable specification language and a set of analysis tools based on static analysis, testing, and model checking. The language enforces an analysis-friendly code structure and offers primitives that allow users to implement their own testers and model checkers directly in the language. We introduce a new approach to static analysis that takes advantage of the capabilities of a rule-based engine. We have applied the analysis-aware methodology to the development of a smart home application.

Radio frequency studies of surface resistance and critical magnetic field of type I and type II superconductors

The surface resistance and the critical magnetic field of lead electroplated on copper were studied at 205 MHz in a half-wave coaxial resonator. The observed surface resistance at a low field level below 4.2°K could be well described by the BCS surface resistance with the addition of a temperature independent residual resistance. The available experimental data suggest that the major fraction of the residual resistance in the present experiment was due to the presence of an oxide layer on the surface. At higher magnetic field levels the surface resistance was found to be enhanced due to surface imperfections.

The attainable rf critical magnetic field between 2.2°K and T_c of lead was found to be limited not by the thermodynamic critical field but rather by the superheating field predicted by the one-dimensional Ginzburg-Landau theory. The observed rf critical field was very close to the expected superheating field, particularly in the higher reduced temperature range, but showed somewhat stronger temperature dependence than the expected superheating field in the lower reduced temperature range.

The rf critical magnetic field was also studied at 90 MHz for pure tin and indium, and for a series of SnIn and InBi alloys spanning both type I and type II superconductivity. The samples were spherical with typical diameters of 1-2 mm and a helical resonator was used to generate the rf magnetic field in the measurement. The results of pure samples of tin and indium showed that a vortex-like nucleation of the normal phase was responsible for the superconducting-to-normal phase transition in the rf field at temperatures up to about 0.98-0.99 T_c' where the ideal superheating limit was being reached. The results of the alloy samples showed that the attainable rf critical fields near T_c were well described by the superheating field predicted by the one-dimensional GL theory in both the type I and type II regimes. The measurement was also made at 300 MHz resulting in no significant change in the rf critical field. Thus it was inferred that the nucleation time of the normal phase, once the critical field was reached, was small compared with the rf period in this frequency range.

Instrumentation for Kinetic-Inductance-Detector-Based Submillimeter Radio Astronomy

A substantial amount of important scientific information is contained within astronomical data at the submillimeter and far-infrared (FIR) wavelengths, including information regarding dusty galaxies, galaxy clusters, and star-forming regions; however, these wavelengths are among the least-explored fields in astronomy because of the technological difficulties involved in such research. Over the past 20 years, considerable efforts have been devoted to developing submillimeter- and millimeter-wavelength astronomical instruments and telescopes.

The number of detectors is an important property of such instruments and is the subject of the current study. Future telescopes will require as many as hundreds of thousands of detectors to meet the necessary requirements in terms of the field of view, scan speed, and resolution. A large pixel count is one benefit of the development of multiplexable detectors that use kinetic inductance detector (KID) technology.

This dissertation presents the development of a KID-based instrument including a portion of the millimeter-wave bandpass filters and all aspects of the readout electronics, which together enabled one of the largest detector counts achieved to date in submillimeter-/millimeter-wavelength imaging arrays: a total of 2304 detectors. The work presented in this dissertation has been implemented in the MUltiwavelength Submillimeter Inductance Camera (MUSIC), a new instrument for the Caltech Submillimeter Observatory (CSO).