Seminars and Defenses
- May 4, 11AM, ENG460
- Yuguang "Michael" Fang • Seminar
- Beef Up the Edge: How to Build a More Powerful IoT System
Connected things in various cyber-physical systems (CPSs), namely IoTs, enable us to sense physical environments, extract intelligent information, and better regulate physical systems we heavily depend on in our daily life. Unfortunately, how to design effective and efficient systems to meet specific applications with diverse quality of service requirements is of paramount importance but highly challenging due to the spatial and temporal variations of user traffic, network spectrum resource, computing capability, storage, and device types. One holistic design approach from the end-to-end perspective seems to be in dire need.
In this talk, the speaker will discuss various related problems and challenges in a connected world and then present a novel collaborative network architecture to enabling connected things to effectively harvest in-network capability (spectrum, energy, storage, and computing power) in a cognitive fashion and intelligently manage the spectrum efficiency, energy efficiency, and yes, security! By pushing in-network capability in communications, computing, and storage to the edge, this network architecture provides an effective and robust approach to IoT.
Dr. Yuguang "Michael" Fang received MS degree from Qufu Normal University, Shandong, China in 1987, PhD degree from Case Western Reserve University in 1994 and PhD degree from Boston University in 1997. He was an assistant professor in Department of Electrical and Computer Engineering at New Jersey Institute of Technology from 1998 to 2000. He then joined the Department of Electrical and Computer Engineering at University of Florida in 2000 and has been a full professor since 2005. He held a University of Florida Research Foundation (UFRF) Professorship (2006-2009, 2017-2020), a University of Florida Term Professorship (2017-2019) and Changjiang Scholar Chair Professorship awarded by the Ministry of Education of China (is currently affiliated with Dalian Maritime University).
Dr. Fang received the US National Science Foundation Career Award in 2001 and the Office of Naval Research Young Investigator Award in 2002, 2015 IEEE Communications Society CISTC Technical Recognition Award, 2014 IEEE Communications Society WTC Recognition Award, and multiple Best Paper Awards from IEEE Globecom (2015, 2011 and 2002) and IEEE ICNP (2006). He has also received 2010-2011 UF Doctoral Dissertation Advisor/Mentoring Award, 2011 Florida Blue Key/UF Homecoming Distinguished Faculty Award, and the 2009 UF College of Engineering Faculty Mentoring Award. He was the Editor-in-Chief of IEEE Transactions on
Vehicular Technology (2013-2017), the Editor-in-Chief of IEEE Wireless Communications (2009-2012), and serves/served on several editorial boards of journals including IEEE Transactions on Mobile Computing (2003-2008, 2011-2016), IEEE Transactions on Communications (2000-2011), and IEEE Transactions on Wireless Communications (2002-2009). He has been actively participating in conference organizations such as serving as the Technical Program Co-Chair for IEEE INFOCOM’2014 and the Technical Program Vice-Chair for IEEE INFOCOM'2005. He is a fellow of the IEEE (2008) and a fellow of the American Association for the Advancement of Science (AAAS) (2015).
- May 2, 10AM, ENG471
- George Seif • MASC thesis defense
- Large Receptive Field Networks for Accurate High-Scale Image Super-Resolution
This thesis presents a novel convolutional neural network architecture for high-scale image super-resolution. In particular, we introduce two separate modifications that can be made to the convolutional layers in the network: one-dimensional separable kernels and dilated kernels. We show how both of these methods can be used to expand the receptive field and performance of super-resolution networks, without increasing the number of trainable parameters or network depth. We show that these modifications can easily be integrated into any convolutional neural network to improve performance. Our methods are especially effective for the challenging high scale super-resolution due to the expanded network receptive field. We conduct extensive empirical evaluations to demonstrate the effectiveness of our methods, showing strong improvements over the state-of-the-art.
- Apr. 25, 2P, ENG460
- Ismail Sheikh • MASC thesis defense
- The Effect of Parallel Execution on Multi-Site Computation Offloading in Mobile Cloud Computing
As smart mobile devices are becoming an indispensable part of our daily life, the demand for running complex applications on such devices are increasing. However, the limitations of resources (e.g. battery life, computation power, and bandwidth) of these devices are restricting the type of applications that can run on them. The restrictions can be overcome by allowing the devices to offload computation and run parts of an application in the powerful cloud servers. The greatest benefit from computation offloading can be obtained by optimally allocating the parts of an application to different devices (i.e. the mobile device and the cloud servers) that minimizes the total cost. The cost can include the response time of the application, mobile battery usage, or the monetary cost incurred by the user for execution on the cloud servers. Further, different types of devices can have a different number of processing cores. Unlike prior work in computation offloading, this work considers
the effect of parallel execution of parts of an application—on different devices (external parallelism) and on the different cores of a single device (internal parallelism)—on offloading allocation. This work models each device as a multi-server queueing station. It uses genetic algorithm to determine the optimal allocation. This work is more advantageous for cases where the workflow of an application has multiple tasks that can be executed in parallel. The results show that considering the effect of parallel execution yields more optimal solution for the allocation problem compared to not accounting for parallelism at all.
- Apr. 10, 12P, ENG460
- Jinming Wen • Seminar
- An Introduction to Integer Signal and Sparse Signal Recovery
In many applications, such as wireless communications, signal processing and GPS, we need to recover an integer parameter vector from an integer linear model. While in some other application domains including computer vision and machine learning, one is frequently required to recover a sparse signal from a few linear measurements. In the first part of this talk, we will theoretically show that some commonly used lattice reductions, which are preprocess tools for recovering integer vectors, can improve the correct recovery probability of some commonly used suboptimal recovery algorithms. In the second part of the talk, we will introduce some algorithms and theory for recovering sparse signals.
Jinming Wen received his Ph.D degree in Applied Mathematics from McGill University in 2015. He was a postdoctoral fellow at ENS de Lyon and University of Alberta from March 2015 to August 2017.He has been a postdoctoral fellow at department of Electrical and Computer Engineering, University of Toronto since September 2017. His research interests include sparse recovery and lattice reduction with applications in communications, signal processing and cryptography. He has published around 30 research papers in top journals and conferences including IEEE Transactions on Information Theory/Signal Processing. He is an Associated Editor of IEEE Access.
- Apr. 12, 10A, ENG471
- Krishnanand Balasundaram • PHD FINAL THESIS DEFENSE
- Morphologically Constrained Adaptive Signal Decomposition in Studying Ventricular Arrhythmias
Ventricular fibrillation (VF), a lethal form of ventricular arrhythmias (VA), originates from the lower chambers of the heart and is one of the major causes for sudden cardiac deaths (SCD). Since the duration from the onset of VF to SCD is only few minutes, it is difficult to study VF and it is even harder to perform invasive diagnosis or provide treatment within the short window of time. This dissertation proposes methods to extract meaningful information from VF electrograms and formulate associations to underlying structural and physiological properties of the cardiac tissue and clinical events of interest during VF. In other words, by analyzing clues in the electrograms during VF, the method can be used to infer the underlying anatomical and physiological properties of the cardiac tissue and certain clinical events of interest, which is otherwise not available without invasive procedures. The proposed methods will be of great assistance to cardiologists and cardiac electrophysiologist in the diagnosis and treatment planning of cardiac arrhythmias.
The proposed adaptive time-frequency (TF) signal decomposition was separated into two categories based on two purposes: (1) Time-specific event detection and (2) Time-averaged VA characterization. For the time-specific event detection (in this work rotor detection because rotors are believed to be drivers of VF), electrogram signal features related to the rotor event were identified with an adaptive TF decomposition and a modified criterion function. Using the proposed features and a LDA based classifier with LOO cross validation, overall classification accuracies of 80.77% and 79.41% were achieved in detecting rotor events and separating them from similar but not rotor events.
In the time-averaged ventricular arrhythmia characterization, previously established signal features were used to associate electrogram clues to the structural and physiological characteristics of the cardiac tissue. Using LCKSVD dictionary learning process, dictionaries of TF basis functions were generated to capture specific electric structures and physiological characteristics of the underlying cardiac tissue. The association of these characteristics with the extracted electrogram clues were validated using a cross-validation technique. The cross-validated results obtained were 71.09%, 60.30%, 71.74% and 65.73%, respectively, for the 4 characteristics.
Further to this, to automate and build a function-approximation model with non-linear separable capabilities to capture complex association of the heart characteristics and electrogram signal structures, neural network models were generated. The cross-validated accuracy for the models were 80.77%, 75.73%, 85.00%, 67.71%, 76.83% and 76.88% respectively for each of the developed models.
- Apr. 4, 11A, ENG460
- Ajmery Sultana • PHD INTERNAL THESIS DEFENSE
- Resource allocation strategies for cognitive radio networks coexistence with D2D communications
Ever increasing demand for enhanced wireless services requires higher data rates, even up to tens of Gbps and pushes the spectrum to its limits. While there is constant push for new spectrum, current licensed spectrum is significantly underutilized. Cognitive radio (CR) approach will provide efficient spectral usage using intelligent wireless nodes. For the successful implementation of cognitive radio networks (CRNs), efficient resource allocation (RA) schemes are essential.
As a first step, the CR approach in radio access networks is introduced. In the second step, the taxonomy of the RA process in CRNs is provided. For radio resource allocation (RRA), the most crucial task is to associate a user with a particular serving base station, to assign the channel and to allocate the power efficiently. In this thesis, a subcarrier assignment scheme and a power allocation algorithm using geometric water-filling (GWF) is presented for orthogonal frequency division multiplexing (OFDM) based CRNs. This algorithm is proved to maximize the sum rate of secondary users by allocating power more efficiently. Then, the resource allocation problem is studied to jointly employ CR technology and device-to-device (D2D) communication in cellular networks in terms of spectral efficiency and energy efficiency (EE). In first case, in terms of spectral efficiency, a two-stage approach is considered to allocate the radio resource efficiently where a new adaptive subcarrier allocation (ASA) scheme is designed first and then a novel power allocation (PA) scheme is developed utilizing proven GWF approach that can compute exact solution with less computation. In second case, in terms of EE, the power allocation problem of cellular networks that co-exist with D2D communication considering both underlay and overlay CR approaches are investigated. A proven power allocation algorithm based on GWF approach are utilized to solve the EE maximization problem which results in an exact and low complexity solution.
- Apr. 3, 12P, ENG460
- Hang Gao • PHD FINAL THESIS DEFENSE
- Model Predictive Control of High Power Current-Source Converter for Medium-Voltage Induction Motor Drive
As a crucial player in medium-voltage (MV) applications, high power current-source converters (CSCs) feature some distinct advantages in contrast to their voltage-source counterparts. However, the traditional control techniques, based on linear proportional-integral (PI) regulators and low band-width modulation, impose several technical issues during low switching frequency operation. In order to meet more and more stringent performance requirements on industrial drives, various high performance finite control-set model predictive control (FCS-MPC) schemes are proposed in this thesis to control CSCs employed in MV induction motor (IM) drives.
The continuous-time and discrete-time dynamic models of high power CSC-fed MV IM drive are developed, which are used to predict the evolution of state variables in the system. Issues related to MPC approach, such as prediction horizon, weighting factor selection, control delay compensation, accurate extrapolation of references, and nature of variable switching frequency are addressed as well.
Model predictive power factor control (MPPFC) is proposed to accurately regulate the line power factor of CSR under various operating conditions. Meanwhile, an active damping function is incorporated into MPPFC to suppress the possible line-side LC resonance. Moreover, an online capacitance estimation method is designed in consideration on the perturbation of the filter parameters of CSR.
In order to keep fixed switching frequency of CSC and improve its dynamic responses, model predictive switching pattern control (MPSPC) and model predictive space vector pattern control (MPSVPC) are proposed, in which MPC technique is combined with selective harmonic elimination (SHE) modulation and space vector modulation (SVM), respectively. In steady state, the PWM waveform of CSC follows the pattern of traditional modulation schemes, whereas during transients CSC is governed by MPC approach for the purpose on dynamic performance improvement.
A common-mode voltage (CMV) reduced model predictive control (RCMV-MPC) is studied, with which the peak value of CMV in high power CSC-fed MV IM drive can be further reduced in comparison with the traditional RCMV modulation schemes. The dynamic responses of the motor drive system are further improved as well.
The simulation on a megawatt motor drive system and experimental results on a low power prototype, validate the effectiveness of the proposed various control schemes.
- Jan. 15, 1P, ENG471
- Jean-Claude Clarke • MASC FINAL THESIS DEFENSE
- Inductive Magnetically Coupled Resonant Wireless Power Transmission System with Planar Coils for the Rezence Efficiency Standard, with a 90% Efficient Class-E Amplifier and Auto-Tuning Circuit Technique for Improved Efficiency
MCR-WPT has helped to increase the WPT distance which the PTU can transmit power efficiently to the PRU. This has enabled their rapid adoption in many applications. In this thesis propose new designs and design methodologies used to meet the Rezence standard for portable electronics, which based on magnetic resonant coupling WPT.
Class E PAs, which achieve 100% theoretical efficiency with simple circuit topology, are difficult to analytically design with equations provided. This thesis presents a HF Class-E PA design methodology that simplifies design. The methodology uses ideal Class-E design equations to generate ideal drain impedances, used to impedance match the power FET output for Class-E operation. A high efficiency Class-E PA was designed using a low cost power MOS by following this design flow.
The small size of these devices makes it difficult to design efficient MCR-WPT resonators. A new multi-layer MCR-WPT PSC design and design methodology is proposed which increases efficiency and decrease SRF of the coils. The design is a modified series stacked spiral inductors without vias, and can be stacked more than 2 layers, and the dielectric substrate changed to a higher permittivity material to improve efficiency and SRF. The design methodology takes advantage of the fast Momentum simulations to give good MCR-WPT system SRF and efficiency predictions.
Efficiency of MCR-WPT Tx and Rx resonators is reduced when the distance between the is less that the CCD due to the frequency splitting phenomenon. This thesis presents an auto-tuning circuit that automatically improve efficiency when the PRU is placed at the optimal region. The tuning is done by a binary counter and discrete capacitor array, and a maximum peak detection technique identifies the optimal tuning capacitance for maximum WPT efficiency.