With fast growing consumer demand for high speed mobile data capacity, wireless spec-trum has become increasingly precious. This drives the evolution of the personal wirelesscommunication, with new standards developed to improve the spectral efficiency. However,the available spectrum below 10GHz is very limited and packing more bits per second into the same bandwidth requires larger energy consumption as well as more stringent radio andMODEM performance. As a result, such an approach is not sustainable for meeting the future demand. A natural path is to move into higher frequency bands which have largerspectrum bandwidth but less commercial usage. Recent years have witnessed vast technology development on V-band (60GHz) Wireless Personal Area Networks (WPAN) andE-band (80GHz) point-to-point cellular backhauls. Meanwhile, the advancement of low-costCMOS technologies enables researchers to significantly improve the integration level of highspeed mm-wave radios with traditional analog and digital circuitry. However, current mmwave radio transmitters suffer from short communication distance and low energy efficiency.This is mainly caused by the reduced performance of the CMOS transmitters employing traditional Power Amplifiers (PAs) that suffer from low transistor breakdown voltage, lowpower gain and poor back-off characteristics. This dissertation investigates the challenges of designing efficient mm-wave transmitters for both long range and short range applications,and proposes concepts and techniques that can potentially break the barriers imposed by the low cost digital CMOS process. The scope of investigation and proposal extends fromthe architecture level down to the transistor level. Specifically, on-chip and spatial power combining techniques are analyzed and implemented to achieve larger transmitter Equiva-lent Isotropically Radiated Power (EIRP). To enhance the average efficiency for modulated signals with high Peak-to-Average-Power-Ratio (PAPR), a direct digital-to-RF conversionarchitecture is proposed and implemented, enabling dynamic DC power scaling. Finally, a Quadrature Spatial Combining concept is introduced to eliminate the tradeoff between lowinsertion loss and high isolation present in a traditional Cartesian architecture with on-chip signal combiners. Prototype chips are fabricated and tested in 65nm CMOS technology toverify the proposed architectures and techniques.

关键词：高级体系结构；无线发射器；CMOS技术

查看摘要 索取原文[148页]

We analyze the s-step biconjugate gradient algorithm in nite precision arithmetic and derive a bound for the residual norm in terms of a minimum polynomial of a perturbed matrixmultiplied by an ampli cation factor. Our bound enables comparison of s-step and classical biconjugate gradient in terms of ampli cation factors. Our results show that for s-step biconjugate gradient, the ampli cation factor depends heavily on the quality of s-step polynomial bases generated in each outer loop.

关键词：双共轭梯度法;有限精度运算;S-级多项式

查看摘要 索取原文[17页]

This thesis explores the design of systems that can climb vertical surfaces with non-negligibledynamics in the sagittal plane. The development of a low-dimensional model addresses a lackof understanding of sagittal-plane dynamics during climbing in the space of reduced-orderdynamic models of legged systems. Using a construction derived from the well-known andwell-studied Spring-Loaded Inverted Pendulum (SLIP), we propose a two-legged system withboth torsional and linear compliance driven by a position-controlled rotational actuator. Twosimple foot models are considered to explore their e?ect on the dynamics and stability of thesystem. Results of the model indicate the existence of passively stable gaits during climbingas well as during inverted running and also suggest mechanical tuning parameters for physicalclimbing systems. A robotic platform capable of producing dynamic climbing behaviors is introduced. A reduced profile, sprawled posture, and improved internal mechanics allow theCLASH platform to be adapted to di?erent climbing substrates. A passive claw engagement mechanism is proposed and tested with simulated steps to verify the design. With thesemechanisms, CLASH becomes the first robotic platform capable of climbing loose cloth and climbs vertically at 15cm/s or 1.5 body-lengths per second. When climbing ferromagneticsurfaces, the system is capable of climbing at 1.8 body-lengths per second. To climb smooth，hard surfaces, a foot with a passively aligning ankle with a tendon-loaded gecko-inspiredadhesive is designed and tested using simulated steps. With these engagement mechanisms，the system is able to climb at 1 body-length per second on acrylic with a 70! incline. Asimple foot-impact model is created to explain the robot’s inability to climb faster or up steeper inclines due to the sagittal-plane reaction forces created during rapid running.

关键词：矢状平面动力学;低维模型;动态模型

查看摘要 索取原文[109页]

Contract-based design is emerging as a unifying compositional paradigm for the speci cation, design and veri cation of large-scale complex systems. Yet, di erent contract frameworks are currently available, without a clear understanding of the relations between them. In this paper, we investigate the relation between interface theories (specifically, relational interfaces) and assume-guarantee (A/G) contracts, revealing some of the subtleties involved. We show that the natural transformation of interfaces to A/G contracts represented by LTL formulas preserves re nement,but does not generally preserve serial composition, and we present an assumption-projection operator to remedy the latter issue. We also discuss the properties of our transformation with respect to conjunction. Finally, we provide illustrative examples that shed light on the e ectiveness of both frameworks for requirement formalization, early detection of integration errors, and principled use of abstraction renement.

关键词：关联接口；担保合同；关系理论；规范化

查看摘要 索取原文[8页]

Distracted driving is still a major concern on roadways today. According to the US Department ofTransportation, in 2011, 387,000 people were injured in crashes involving a distracted driver.While there has been tremendous work in vehicular safety systems, as the velocity of the vehicleincreases, threat assessment becomes more difficult to predict. In this regime, understanding howthe driver behaves is critical in determining the safety of the vehicle. This work presents aframework for incorporating human driver modeling with a controller to provide asemi-autonomous system for vehicle safety. We describe how the system is built, present metricsto evaluate its utility and conduct real-time experiments with a safety controller to verify theutility of incorporating the driver’s behavior while designing a semi-autonomous system.

关键词：驱动程序;半自治控制;建模;安全系统

查看摘要 索取原文[54页]

The empirical observation of human locomotion has found considerable utility in the diagnosisof numerous neuromuscular pathologies. Unfortunately without the construction of a dynamical system model of the measured gait, the e ectualness of these observations isrestricted to just the existing diagnostic variety rather than the prediction of potential instabilities in gait or guiding the construction of user speci c prosthetics. In order to construct adynamical system model of an observed gait in an automated fashion, one requires a familyof representations rich enough to describe the dynamics of gait and an automated procedureto select a particular representation capable of describing a given observation from thisfamily.

关键词：运动识别;混合动力模型;切换系统;最优控制

查看摘要 索取原文[142页]

In this report, we analyze the computational and performance aspects of current stateof the-art object recognition and optical flow algorithms. First, we identify important algorithms for object recognition and optical flow, then we perform a pattern decomposition to identify key computations. We include pro les of the runtime and energy eciency (GFLOPS/W) for our implementation of these applications on a commercial architecture.Finally, we include an analysis of memory-bandwidth boundedness for optical flow to identify opportunities for communication-avoiding algorithms.

关键词：对象识别；光学流；能源效率；有界性分析

查看摘要 索取原文[22页]

FMI (Functional Mockup Interface) is a standard for exchanging and co-simulating model components (called FMUs) coming from potentially di erent modeling formalisms, languages, and tools. Previous work has proposed a formal model for the co-simulation part of the FMI standard, and also presented two co-simulation algorithms which can be proven to have desirable properties, such as determinacy, provided the FMUs satisfy a formal contract. In this paper we discuss the principles for encoding di erent modeling formalisms, including state machines, discrete-event systems, and synchronous data ow, as FMUs. The challenge is to bridge the various semantic gaps (untimed vs. timed, signals vs. events, etc.) that arise because of the heterogeneity between these modeling formalisms and the FMI API.

关键词：桥接异构；功能模型接口；FMI

查看摘要 索取原文[23页]

Field Programmable Gate Arrays (FPGAs) are effective in satisfying different computingrequirements such as high performance computing, digital signal processingand embedded computing. Academic research efforts cover architecture optimization,design space exploration, circuit level design and new CAD algorithms. However,actual FPGA devices in common use are all commercial. In this work, we designedand implemented an open source FPGA with toolflow support. We named the project Archipelago and designed the architecture with two goals. First, it explores the qualityof the physical implementation result that is produced by a standard ASIC design flow in a modern ASIC process. Second, it enables other people to use and extend theproject at will. The outcome of this project is a parameterizable and user expandable FPGA with toolflow support. We verified its functionality. The performance of thework is good enough for real world work loads. A 64 Bit counter can run up to 364MHz on average on three different FPGA instances. The result is comparable to acommercial FPGA implemented in a similar process technology.

关键词：可编程门阵列(fpga)；高性能计算；数字信号处理；嵌入式计算

查看摘要 索取原文[79页]

Printed electronics employing solution-processed materials is considered to be the key torealizing low-cost large-area electronic systems, but the performance of printed transistors isnot generally adequate for most intended applications due to limited performance of printablesemiconductors. In this dissertation, I propose an alternative approach of a printedswitch, where the use of semiconductors can be avoided by building mechanical switcheswith printed metal nanoparticles. I provide the rst demonstration of inkjet-printed microelectro-mechanical (MEM) switches with abrupt switching characteristics, very low on-stateresistance ( 10), and nearly perfect o -state behavior with immeasurable leakage withon/o current ratio of 107. The devices are fabricated using a novel process scheme to build3-dimensional cantilever structures from solution-processed metallic nanoparticles and sacricial polymers. These printed MEM switches thus represent a uniquely attractive path forrealizing printed electronics. I will also discuss an inkjet-printed microshell encapsulation asa new zero-level packaging technology. Inkjet-printing of silver nanoparticle ink is demonstratedto form porous microshells through which sacri cial oxide can be selectively removedto release MEMS structures. A second inkjet printing process using ner gold nanoparticleink or polymer is demonstrated to e ectively seal the microshells. This inkjet-printed microshell encapsulation technology is successfully applied to a MEM relay, and is demonstratedto mitigate the issue of contact oxidation. Speci cally, the stability of the relayON-state resistance is dramatically improved by more than a factor of 100.

关键词：微机电系统；印刷晶体管；半导体

查看摘要 索取原文[101页]