Search Content

Due to the reduced fuel usage and greenhouse emission advantage, the sales of electric vehicles (EV) have risen dramatically in recent years. Generally speaking, the EVs are pursuing higher power and lighter weight, which requires higher power density for all the power electronics converters in the EVs. To design higher…

Due to the reduced fuel usage and greenhouse emission advantage, the sales of electric vehicles (EV) have risen dramatically in recent years. Generally speaking, the EVs are pursuing higher power and lighter weight, which requires higher power density for all the power electronics converters in the EVs. To design higher density power converters, three key emerging power electronics technologies are investigated in this study. First, the PCB-based magnetics are beneficial for improving the power density due to their low-profile structure. However, the high winding capacitance is considered one of the significant drawbacks of PCB-based magnetics. In this study, a novel winding structure is proposed to cut down the winding capacitance by 75% with little compromise of the winding loss. Second, the synchronous rectifiers (SR) are usually utilized to improve the system efficiency and power density compared with the conventional diode bridge rectifiers for the AC/DC stage in the power converters. The SRs are desired to be turned off at current zero-crossing to generate a minimal loss. However, the precise current zero-crossing detection is very challenging in high-frequency and high-power-density converters. In this study, a high-dv/dt-immune and parameter-adaptive SR driving scheme is proposed to guarantee the zero-current switching (ZCS) of SRs in various operating conditions and improve the system efficiency by 1.23%. Finally, Gallium Nitride (GaN) semiconductors are considered less lossy than Silicon (Si) semiconductors. However, the voltage rating of the commercial GaN HEMTs is limited to 600/650 V due to the lateral structure, which is not suitable for the 800 V or higher dc-link voltage EV systems. Stacking the low-voltage rating devices is a straightforward approach to sustain higher dc-link voltage. However, unbalanced voltage sharing can occur, which can damage the low-voltage rating devices in the stack. In this study, a novel active current source gate driver is proposed to suppress the over-voltage of the stacking devices below 10% for all operating conditions without sacrificing switching speed or switching energy. The above emerging power electronics technologies are investigated thoroughly in the dissertation. The proposed approaches are practical for improving power converters’ density in future EV applications.

ContributorsZhang, Zhengda (Author) / Lei, Qin Q.L. (Thesis advisor) / Ayyanar, Raja R.A. (Committee member) / Yu, Hongbin H.Y. (Committee member) / Pal, Anamitra A.P. (Committee member) / Ranjram, Mike M.R. (Committee member) / Arizona State University (Publisher)

Created2021

Set-Valued Methods for Reachability Analysis and Estimation of Nonlinear Dynamical Systems

Description

The goal of this thesis research is to contribute to the design of set-valued methods, i.e., algorithms that leverage a set-theoretic framework that can provide a powerful means for control designs for general classes of uncertain nonlinear dynamical systems, and in particular, to develop set-valued algorithms for constrained reachability problems…

The goal of this thesis research is to contribute to the design of set-valued methods, i.e., algorithms that leverage a set-theoretic framework that can provide a powerful means for control designs for general classes of uncertain nonlinear dynamical systems, and in particular, to develop set-valued algorithms for constrained reachability problems and estimation.I propose novel fixed-order hyperball-valued observers for different classes of nonlinear systems, including Linear Parameter Varying, Lipschitz continuous and Decremental Quadratic Constrained nonlinearities, with unknown inputs that simultaneously find bounded sets of states and unknown inputs that contain the true states and inputs and are compatible with the measurement/outputs. In addition, I provide sufficient conditions for the existence and stability of the estimates, the convergence of the estimation errors, and the optimality of the observers. Moreover, I design state and unknown input observers, as well as mode detectors for hidden mode, switched linear and nonlinear systems with bounded-norm noise and unknown inputs. To address this, I propose a multiple-model approach to obtain a bank of mode-matched set-valued observers in combination with a novel mode observer, based on elimination. My mode elimination approach uses the upper bound of the norm of to-be-designed residual signals to remove inconsistent modes from the bank of observers. I also provide sufficient conditions for mode detectability. Furthermore, I address the problem of designing interval observers for partially unknown nonlinear systems, using affine abstractions, nonlinear decomposition functions, and a data-driven function over-approximation approach to over-estimate the unknown dynamic model. The proposed observer recursively computes the correct interval estimates. Then, using observed measurement signals, the observer iteratively shrinks the intervals. Moreover, the observer updates the over-approximation model of the unknown dynamics. Finally, I propose a tractable family of remainder-from decomposition functions for a broad range of dynamical systems. Moreover, I introduce a set-inversion algorithm that along with the proposed decomposition functions have several applications, e.g., in the approximation of the reachable sets for bounded-error, constrained, continuous, and/or discrete-time systems, as well as in guaranteed state estimation. Leveraging mixed-monotonicity, I provide novel set-theoretic approaches to address the problem of polytope-valued state estimation in bounded-error discrete-time nonlinear systems, subject to nonlinear observations/constraints.

ContributorsKhajenejad, Mohammad (Author) / Zheng Yong, Sze S.Z.Y (Thesis advisor) / Nedich, Angelia A.N (Committee member) / Reffett, Kevin K.R (Committee member) / M. Berman, Spring S.M.B (Committee member) / Fainekos, Georgios G.F (Committee member) / Lee, Hyunglae H.L (Committee member) / Arizona State University (Publisher)

Created2021

Distributed Learning and Adaptive Algorithms for Edge Networks

Description

Edge networks pose unique challenges for machine learning and network management. The primary objective of this dissertation is to study deep learning and adaptive control aspects of edge networks and to address some of the unique challenges therein. This dissertation explores four particular problems of interest at the intersection of…

Edge networks pose unique challenges for machine learning and network management. The primary objective of this dissertation is to study deep learning and adaptive control aspects of edge networks and to address some of the unique challenges therein. This dissertation explores four particular problems of interest at the intersection of edge intelligence, deep learning and network management. The first problem explores the learning of generative models in edge learning setting. Since the learning tasks in similar environments share model similarity, it is plausible to leverage pre-trained generative models from other edge nodes. Appealing to optimal transport theory tailored towards Wasserstein-1 generative adversarial networks, this part aims to develop a framework which systematically optimizes the generative model learning performance using local data at the edge node while exploiting the adaptive coalescence of pre-trained generative models from other nodes. In the second part, a many-to-one wireless architecture for federated learning at the network edge, where multiple edge devices collaboratively train a model using local data, is considered. The unreliable nature of wireless connectivity, togetherwith the constraints in computing resources at edge devices, dictates that the local updates at edge devices should be carefully crafted and compressed to match the wireless communication resources available and should work in concert with the receiver. Therefore, a Stochastic Gradient Descent based bandlimited coordinate descent algorithm is designed for such settings. The third part explores the adaptive traffic engineering algorithms in a dynamic network environment. The ages of traffic measurements exhibit significant variation due to asynchronization and random communication delays between routers and controllers. Inspired by the software defined networking architecture, a controller-assisted distributed routing scheme with recursive link weight reconfigurations, accounting for the impact of measurement ages and routing instability, is devised. The final part focuses on developing a federated learning based framework for traffic reshaping of electric vehicle (EV) charging. The absence of private EV owner information and scattered EV charging data among charging stations motivates the utilization of a federated learning approach. Federated learning algorithms are devised to minimize peak EV charging demand both spatially and temporarily, while maximizing the charging station profit.

ContributorsDedeoglu, Mehmet (Author) / Zhang, Junshan (Thesis advisor) / Kosut, Oliver (Committee member) / Zhang, Yanchao (Committee member) / Fan, Deliang (Committee member) / Arizona State University (Publisher)

Created2021

Meta-Learning in Edge Networks: Model-Based Reinforcement Learning and Distributed Edge Learning

Description

Pushing the artificial intelligence frontier to resource-constrained edge nodes for edge intelligence is nontrivial. This dissertation provides a comprehensive study of optimization-based meta-learning algorithms to build a theoretic foundation of edge intelligence, with the focus on two topics: 1) model-based reinforcement learning (RL); 2) distributed edge learning. Under this common…

Pushing the artificial intelligence frontier to resource-constrained edge nodes for edge intelligence is nontrivial. This dissertation provides a comprehensive study of optimization-based meta-learning algorithms to build a theoretic foundation of edge intelligence, with the focus on two topics: 1) model-based reinforcement learning (RL); 2) distributed edge learning. Under this common theme, this study is broadly organized into two parts. The first part studies meta-learning algorithms for model-based RL. First, the fundamental limit of model learning is explored for linear time-varying systems, using a two-step meta-learning algorithm with an episodic block model. A comprehensive non-asymptotic analysis of the sample complexity is provided, where a two-scale martingale small-ball approach is devised to address the challenges in sample correlation and small sample sizes. Next, policy learning of offline RL in general Markov decision processes is explored. To tackle the challenges therein, e.g., value overestimation and possibly poor quality of offline datasets, a model-based offline Meta-RL approach with regularized policy optimization is proposed, by learning a meta-model for task inference and a meta-policy for safe exploration of out-of-distribution state-actions. The second part investigates meta-learning algorithms for distributed edge learning. First, the general edge supervised learning is considered, where the edge node aims to quickly learn a good model with limited samples. A platform-aided collaborative learning framework is proposed to learn a model initialization via federated meta-learning across multiple nodes, which is transferred to target nodes for fine-tuning. Then, a channel gating module is introduced to select important channels of backbone networks for efficient local computation. A novel federated meta-learning approach is developed to learn meta-initializations for backbone networks and gating modules, from which a task-specific channel gated network is quickly adapted. Taking one step further, the continual edge learning is investigated in the context of online meta-learning, where each node has a sequence of online tasks. A multi-agent online meta-learning framework is developed to accelerate the task-average performance in a single node under limited communication among neighbors, through the lens of distributed online convex optimization. Building on distributed online gradient descent with gradient tracking, the optimal task-average regret is achieved at a faster rate.

ContributorsLin, Sen (Author) / Zhang, Junshan JZ (Thesis advisor) / Ying, Lei LY (Thesis advisor) / Bertsekas, Dimitri DB (Committee member) / Nedich, Angelia AN (Committee member) / Wang, Weina WW (Committee member) / Arizona State University (Publisher)

Created2021

Intelligent Phase Shifter for mm-wave Beam Forming.

Description

Modern communication systems call for state-of-the-art links that offer almost idealistic performance. This requirement had pushed the technological world to pursue communication in frequency bands that were almost incomprehensible back when the first series of cordless cellphones were invented. These requirements have impacted everything from civilian requirements, space, medical diagnostics…

Modern communication systems call for state-of-the-art links that offer almost idealistic performance. This requirement had pushed the technological world to pursue communication in frequency bands that were almost incomprehensible back when the first series of cordless cellphones were invented. These requirements have impacted everything from civilian requirements, space, medical diagnostics to defense technologies and have ushered in a new era of advancements. This work presents a new and novel approach towards improving the conventional phased array systems. The Intelligent Phase Shifter (IPS) offers phase tracking and discrimination solutions that currently plague High-Frequency wireless systems. The proposed system is implemented on (CMOS) process node to better scalability and reduce the overall power dissipated. A tracking system can discern Radio Frequency (RF) Signals’ phase characteristics using a double-balanced mixer. A locally generated reference signal is then matched to the phase of the incoming receiver using a fully modular yet continuous complete 360ᵒ phase shifter that alters the phase of the local reference and matches the phase with that of an incoming RF reference. The tracking is generally two control voltages that carry In-phase and Quadrature-phase information. These control signals offer the capability of controlling similar devices when placed in an array and eliminating any ambiguity that might occur due to in-band interference.

ContributorsLakshminarasimhaiah Rajendra, Yashas (Author) / Zeinolabedinzadeh, Saeed (Thesis advisor) / Trichopoulos, Georgios (Committee member) / Aberle, James (Committee member) / Arizona State University (Publisher)

Created2021

Filtering by