An extended complex system is a large scale hierarchical system controlled by two or more teams of decision-makers. The teams may have different objective functions, and often can be in direct conflict with each other. Within each team, the decision-makers must cooperate for the collective benefit of the team, but outside the team each member must compete with the decision-makers in the other teams. Decision-making in the context of such an extended complex system can be modeled as a new framework within the theory of games, called multi-team games. A multi-team game is a decision-making structure consisting of several interacting teams of cooperating decision-makers that are simultaneously in conflict with the other teams. In this dissertation, a new strategy, called Noninferior Nash strategy, is proposed to define optimal cooperative decisions for members of non-cooperative teams in an extended complex system. This strategy represents an equilibrium for the teams characterized by the property that no team has an incentive to unilaterally deviate, while maintaining cooperation among its members, in order to improve its overall team performance. The Noninferior Nash strategy in both static and dynamic systems is developed and its properties are investigated. In order to deal with the issue of non-uniqueness of the solution, a team structure that allows for a leader to oversee the overall performance of the team is introduced. The Noninferior Nash strategy with a Leader is formulated so as to select the particular Noninferior Nash strategy that is best for the team. To illustrate these concepts on a realistic system, we consider a practical example of a military air operation modeled as an extended complex system. The Nash Noninferior Strategies are investigated as possible solution concepts for dynamic teaming, team tasking, and unit task assignments and reassignments in the process of optimally planning of shared responsibilities and roles in the hierarchical deployment of the units in the combat. Simulation examples are presented to illustrate the effectiveness of these strategies in preserving the friendly force while destroying the defending enemy units.

To be added later...

Silicon Germanium (Si1-xGex) is considered the choice for analog/mixed-signal RF and optoelectronic systems due to its high speed, low noise and compatibility with standard CMOS processes. The goal of this thesis is to investigate photo-detection in SiGe and optical receiver circuits in the commercially available IBM 5HP SiGe BiCMOS process. The study of photodetectors based on SiGe is of interest because of its high absorption capability at wavelengths between 1.1-1.5um. In this thesis several designs of receiver circuits and front-end transimpedance amplifiers (TIA) were designed and fabricated in the IBM 5HP (0.5µm) SiGe technology exhibiting high transit (fT) and maximum oscillation frequency (fMAX). Spectre simulations for both the transimpedance amplifiers and the complete receiver circuits are conducted at the single supply voltage of 3.3V in the Cadence Analog Affirma design environment. The analog mixed signal design tools NeoCircuit/NeoCell from Neolinear Inc. and Analog Affirma from Cadence Inc. are used for the optimization of the complete receiver circuits consisting of a transimpedance amplifier, a cascaded multi-stage differential amplifier and a decision circuit.

The architecture of modern FPGAs contain over one thousand 512-bit memory banks, over five hundred 4k-bit memory banks, and over one hundred thousand logic elements. This inherent parallelism of an FPGA makes it an ideal platform for a multiprocessor architecture. In addition to embedded memory, hundreds of ASIC multipliers are embedded into modern FPGA architectures. This thesis introduces three Single-Instruction-Multiple-Data architectures comprised of 2, 4, 8, 16, 32, 64 and 88 processing elements. The first architecture uses configurable logic to implement the processing elements while second and third architectures are built around ASIC multipliers and use configurable logic to implement customizable instruction. All of the architectures described in this thesis are controlled by a central instruction stream. The 64 interconnected processor SIMD design operates at 94 MHz, and utilizes 73% of the DSP blocks available in the Altera Stratix EPS80F1508C6 device but only 24% of the look-up table logic. The remaining 76% of the logic cells are available for custom instructions.

This thesis presents a system-level co-simulation environment for mixed domain design exploration. By employing shared memory IPC (Inter-Process Communication) and utilizing PDES (Parallel Discrete Event Simulation) techniques, we examine two methods of synchronization, lock-step and dynamic. We then compare the performance of these two methods on a series of test systems as well as real designs using the Chatoyant MOEMS (Micro-Electro Mechanical Systems) simulator and the mixed HDL (Hardware Description Language) simulator from Model Technology, ModelSim. The results collected are used to ascertain which method provides the best overall performance with the least overhead.

The IEEE 1451 standards define sets of common communication interfaces to standardize the connectivity of transducer to microprocessor, instrumentation systems, and networks. This thesis presents a single chip solution for these standards using Alteras Excalibur chip. All pertinent elements of the standard for the Transducer Interface Model (TIM) and Network Capable Application Processor (NCAP) were implemented using a combination of hardware (FPGA) and software. The total design takes advantage of the architecture provided by the Excalibur chip. Past solutions have only implemented the TIM and interfaced it to an NCAP using the digital communication protocol defined in the standard. This solution eliminates inter-chip communication. Rather, an on-chip parallel connection is established between the TIM and NCAP through a high performance bus.

Government regulations require employers to monitor for harmful chemicals in the work place. Ethylene oxide is one such chemical, but a large number of other chemicals must also be monitored. Existing monitoring systems are targeted at a small range of chemicals or a single chemical. A system to monitor for a new chemical often requires development of a new system. This is expensive and increases the cost of the system. Reuse technology, currently used in numerous areas for rapid product development and for minimizing cost should be applied to the hazardous gas detection field. This thesis describes the development and implementation of software for a distributed system that controls a hazardous gas monitoring system capable of interfacing with multiple types of sensors. The prototype system was targeted at detection of ethylene oxide. Requirements were developed after reviewing existing monitoring systems, and determining desired additional functionality and reuse requirements. Specifications followed from the requirements. Software was designed from the specifications and then implemented and tested.

Fabricating semiconductor junction diodes can be costly in price and time. It typically involves long turn around time from the foundry, which requires precise machines and a clean room environment to prevent contamination from air born materials. This intensive process of fabricating diodes, means the cost will increase as the machines to process the semiconductor material need to become more precise. The proposed solution to the problem of cost and time is to create a printing system that is capable of printing diodes and circuits on various substrates. The experiments described in this thesis investigate the possible diode material, Poly(3,4-ethylenedioxythiophene), a polymer in aqueous dispersion. Variation in experiments with the weight ratios for the chemicals used for synthesis of this polymer are used as a benchmark to find the optimum weight ratios for the Poly(3,4-ethylenedioxythiophene) synthesis, to create an asymmetric rectifying diode, using continuous ink jet printing technology.

It seems these days that everyone has a cellular phone. Whether yours is for business purposes or personal use, you need an efficient way of charging the battery in the phone. But, like most people, you probably dont like being tethered to the wall. Imagine a system where your cellular phone battery is always charged. No more worrying about forgetting to charge the battery. Sound Impossible? It is the focus of this thesis to discuss the first step toward realizing this goal. A system will be presented using existing antenna and charge pump technology to charge a cellular phone battery without wires. In this first step, we will use a standard phone, and incorporate the charging technology into a commercially available base station. The base station will contain an antenna tuned to 915MHz and a charge pump. We will discuss the advantages and disadvantages of such a system, and hopefully pave the way for a system incorporated into the phone for charging without the use of a base station.

The capability of artificial neural networks for pattern recognition of real world problems is well known. In recent years, the support vector machine has been advocated for its structure risk minimization leading to tolerance margins of decision boundaries. Structures and performances of these pattern classifiers depend on the feature dimension and training data size. The objective of this research is to compare these pattern recognition systems based on a case study. The particular case considered is on classification of hypertensive and normotensive right ventricle (RV) shapes obtained from Magnetic Resonance Image (MRI) sequences. In this case, the feature dimension is reasonable, but the available training data set is small, however, the decision surface is highly nonlinear. For diagnosis of congenital heart defects, especially those associated with pressure and volume overload problems, a reliable pattern classifier for determining right ventricle function is needed. RV¡¦s global and regional surface to volume ratios are assessed from an individual¡¦s MRI heart images. These are used as features for pattern classifiers. We considered first two linear classification methods: the Fisher linear discriminant and the linear classifier trained by the Ho-Kayshap algorithm. When the data are not linearly separable, artificial neural networks with back-propagation training and radial basis function networks were then considered, providing nonlinear decision surfaces. Thirdly, a support vector machine was trained which gives tolerance margins on both sides of the decision surface. We have found in this case study that the back-propagation training of an artificial neural network depends heavily on the selection of initial weights, even though randomized. The support vector machine where radial basis function kernels are used is easily trained and provides decision tolerance margins, in spite of only small margins.