Search

Flexible ramping products (FRPs) emerge as a promising instrument for addressing steep and uncertain ramping needs through market mechanisms. Initial implementations of FRPs in North American electricity markets, however, revealed several shortcomings in existing FRP designs. In many instances, FRP prices failed to signal the true value of ramping capacity, most notably evident in zero FRP prices observed in a myriad of periods during which the system was in acute need for rampable capacity. These periods were marked by scheduled but undeliverable FRPs, often calling for operator out-of-market actions. On top of that, the methods used for procuring FRPs have been primarily rule-based, lacking explicit economic underpinnings. In this paper, we put forth an alternative framework for FRP procurement, which seeks to set FRP requirements and schedule FRP awards such that the expected system operation cost is minimized. Using real world data from U.S. ISOs, we showcase the relative merits of the framework in (i) reducing the total system operation cost, (ii) improving price formation, (iii) enhancing the the deliverability of FRP awards, and (iv) reducing the need for out-of-market actions.

In this work, we propose a new method to track extended targets of different shapes such as ellipses, rectangles and rhombi. We provide an analytical framework to express these shapes as superelliptical contours and propose a Bayesian filtering scheme that can handle measurements from the contour of the object. The method utilizes the Rao-Blackwellized particle filtering algorithm with novel sensor-object geometry constraints. The success of the algorithm is demonstrated using both simulations and real-data experiments, and the algorithm has been demonstrated to be of high performance in various challenging scenarios., Comment: 10 pages, 6 figures, accepted by the 27th International Conference on Information Fusion (FUSION 2024)

This paper investigates the potential of Digital Twins (DTs) to enhance network performance in densely populated urban areas, specifically focusing on vehicular networks. The study comprises two phases. In Phase I, we utilize traffic data and AI clustering to identify critical locations, particularly in crowded urban areas with high accident rates. In Phase II, we evaluate the advantages of twinning vehicular networks through three deployment scenarios: edge-based twin, cloud-based twin, and hybrid-based twin. Our analysis demonstrates that twinning significantly reduces network delays, with virtual twins outperforming physical networks. Virtual twins maintain low delays even with increased vehicle density, such as 15.05 seconds for 300 vehicles. Moreover, they exhibit faster computational speeds, with cloud-based twins being 1.7 times faster than edge twins in certain scenarios. These findings provide insights for efficient vehicular communication and underscore the potential of virtual twins in enhancing vehicular networks in crowded areas while emphasizing the importance of considering real-world factors when making deployment decisions., Comment: 6 pages, 8 figures, 2tables, conference paper

Detecting machine malfunctions at an early stage is crucial for reducing interruptions in operational processes within industrial settings. Recently, the deep learning approach has started to be preferred for the detection of failures in machines. Deep learning provides an effective solution in fault detection processes thanks to automatic feature extraction. In this study, a deep learning-based system was designed to analyze the sound signals produced by industrial machines. Acoustic sound signals were converted into Mel spectrograms. For the purpose of classifying spectrogram images, the DenseNet-169 model, a deep learning architecture recognized for its effectiveness in image classification tasks, was used. The model was trained using the transfer learning method on the MIMII dataset including sounds from four types of industrial machines. The results showed that the proposed method reached an accuracy rate varying between 97.17% and 99.87% at different Sound Noise Rate levels.

The dramatic increase in the connectivity demand results in an excessive amount of Internet of Things (IoT) sensors. To meet the management needs of these large-scale networks, such as accurate monitoring and learning capabilities, Digital Twin (DT) is the key enabler. However, current attempts regarding DT implementations remain insufficient due to the perpetual connectivity requirements of IoT networks. Furthermore, the sensor data streaming in IoT networks cause higher processing time than traditional methods. In addition to these, the current intelligent mechanisms cannot perform well due to the spatiotemporal changes in the implemented IoT network scenario. To handle these challenges, we propose a DT-native AI-driven service architecture in support of the concept of IoT networks. Within the proposed DT-native architecture, we implement a TCP-based data flow pipeline and a Reinforcement Learning (RL)-based learner model. We apply the proposed architecture to one of the broad concepts of IoT networks, the Internet of Vehicles (IoV). We measure the efficiency of our proposed architecture and note ~30% processing time-saving thanks to the TCP-based data flow pipeline. Moreover, we test the performance of the learner model by applying several learning rate combinations for actor and critic networks and highlight the most successive model.

As global Internet of Things (IoT) devices connectivity surges, a significant portion gravitates towards the Edge of Things (EoT) network. This shift prompts businesses to deploy infrastructure closer to end-users, enhancing accessibility. However, the growing EoT network expands the attack surface, necessitating robust and proactive security measures. Traditional solutions fall short against dynamic EoT threats, highlighting the need for proactive and intelligent systems. We introduce a digital twin-empowered smart attack detection system for 6G EoT networks. Leveraging digital twin and edge computing, it monitors and simulates physical assets in real time, enhancing security. An online learning module in the proposed system optimizes the network performance. Our system excels in proactive threat detection, ensuring 6G EoT network security. The performance evaluations demonstrate its effectiveness, robustness, and adaptability using real datasets.

Deep edge intelligence aims to deploy deep learning models that demand computationally expensive training in the edge network with limited computational power. Moreover, many deep edge intelligence applications require handling distributed data that cannot be transferred to a central server due to privacy concerns. Decentralized learning methods, such as federated learning, offer solutions where models are learned collectively by exchanging learned weights. However, they often require complex models that edge devices may not handle and multiple rounds of network communication to achieve state-of-the-art performances. This study proposes a convolutional ensemble learning approach, coined EdgeConvEns, that facilitates training heterogeneous weak models on edge and learning to ensemble them where data on edge are heterogeneously distributed. Edge models are implemented and trained independently on Field-Programmable Gate Array (FPGA) devices with various computational capacities. Learned data representations are transferred to a central server where the ensemble model is trained with the learned features received from the edge devices to boost the overall prediction performance. Extensive experiments demonstrate that the EdgeConvEns can outperform the state-of-the-art performance with fewer communications and less data in various training scenarios.

In deep learning inference, model parameters are pruned and quantized to reduce the model size. Compression methods and common subexpression (CSE) elimination algorithms are applied on sparse constant matrices to deploy the models on low-cost embedded devices. However, the state-of-the-art CSE elimination methods do not scale well for handling large matrices. They reach hours for extracting CSEs in a $200 \times 200$ matrix while their matrix multiplication algorithms execute longer than the conventional matrix multiplication methods. Besides, there exist no compression methods for matrices utilizing CSEs. As a remedy to this problem, a random search-based algorithm is proposed in this paper to extract CSEs in the column pairs of a constant matrix. It produces an adder tree for a $1000 \times 1000$ matrix in a minute. To compress the adder tree, this paper presents a compression format by extending the Compressed Sparse Row (CSR) to include CSEs. While compression rates of more than $50\%$ can be achieved compared to the original CSR format, simulations for a single-core embedded system show that the matrix multiplication execution time can be reduced by $20\%$.

Optical coherence tomography (OCT) is a label-free, non-invasive 3D imaging tool widely used in both biological research and clinical diagnosis. Current OCT modalities can only visualize specimen tomography without chemical information. Here, we report a bondselective full-field OCT (BS-FF-OCT), in which a pulsed mid-infrared laser is used to modulate the OCT signal through the photothermal effect, achieving label-free bond-selective 3D sectioned imaging of highly scattering samples. We first demonstrate BS-FF-OCT imaging of 1 {\mu}m PMMA beads embedded in agarose gel. Next, we then show 3D hyperspectral imaging of polypropylene fiber mattress from a standard surgical mask. We then demonstrate BS-FFOCT imaging on biological samples, including cancer cell spheroids and C. elegans. Using an alternative pulse timing configuration, we finally demonstrate the capability of BS-FF-OCT on a bulky and highly scattering 150 {\mu}m thick mouse brain slice.

Clinical identification and fundamental study of viruses rely on the detection of viral proteins or viral nucleic acids. Yet, amplification-based and antigen-based methods are not able to provide precise compositional information of individual virions due to small particle size and low-abundance chemical contents (e.g., ~ 5000 proteins in a vesicular stomatitis virus). Here, we report a widefield interferometric defocus-enhanced mid-infrared photothermal (WIDE-MIP) microscope for high-throughput fingerprinting of single viruses. With the identification of feature absorption peaks, WIDE-MIP reveals the contents of viral proteins and nucleic acids in single DNA vaccinia viruses and RNA vesicular stomatitis viruses. Different nucleic acids signatures of thymine and uracil residue vibrations are obtained to differentiate DNA and RNA viruses. WIDE-MIP imaging further reveals an enriched \b{eta} sheet components in DNA varicella-zoster virus proteins. Together, these advances open a new avenue for compositional analysis of viral vectors and elucidating protein function in an assembled virion.

Storage is expected to be a critical source of firming in low-carbon grids. A common concern raised from ex-post assessments is that storage resources can fail to respond to strong price signals during times of scarcity. While commonly attributed to forecast error or failures in operations, we posit that this behavior can be explained from the perspective of risk-averse scheduling. Using a stochastic self-scheduling framework and real-world data harvested from the Australian National Electricity Market, we demonstrate that risk-averse storage resources tend to have a myopic operational perspective, that is, they typically engage in near-term price arbitrage and chase only few extreme price spikes and troughs, thus remaining idle in several time periods with markedly high and low prices. This has important policy implications given the non-transparency of unit risk aversion and apparent risk in intervention decision-making., Comment: 5 pages

To take unit commitment (UC) decisions under uncertain net load, most studies utilize a stochastic UC (SUC) model that adopts a one-size-fits-all representation of uncertainty. Disregarding contextual information such as weather forecasts and temporal information, these models are typically plagued by a poor out-of-sample performance. To effectively exploit contextual information, in this paper, we formulate a conditional SUC problem that is solved given a covariate observation. The presented problem relies on the true conditional distribution of net load and so cannot be solved in practice. To approximate its solution, we put forward a predictive prescription framework, which leverages a machine learning model to derive weights that are used in solving a reweighted sample average approximation problem. In contrast with existing predictive prescription frameworks, we manipulate the weights that the learning model delivers based on the specific dataset, present a method to select pertinent covariates, and tune the hyperparameters of the framework based on the out-of-sample cost of its policies. We conduct extensive numerical studies, which lay out the relative merits of the framework vis-\`a-vis various benchmarks., Comment: 5 pages, 2 figures

Safely navigating through an urban environment without violating any traffic rules is a crucial performance target for reliable autonomous driving. In this paper, we present a Reinforcement Learning (RL) based methodology to DEtect and FIX (DeFIX) failures of an Imitation Learning (IL) agent by extracting infraction spots and re-constructing mini-scenarios on these infraction areas to train an RL agent for fixing the shortcomings of the IL approach. DeFIX is a continuous learning framework, where extraction of failure scenarios and training of RL agents are executed in an infinite loop. After each new policy is trained and added to the library of policies, a policy classifier method effectively decides on which policy to activate at each step during the evaluation. It is demonstrated that even with only one RL agent trained on failure scenario of an IL agent, DeFIX method is either competitive or does outperform state-of-the-art IL and RL based autonomous urban driving benchmarks. We trained and validated our approach on the most challenging map (Town05) of CARLA simulator which involves complex, realistic, and adversarial driving scenarios. The source code is publicly available at https://github.com/data-and-decision-lab/DeFIX, Comment: 6 pages, 4 figures, 2 tables, published in IEEE International Conference on Intelligent Transportation Systems (ITSC), October 12, 2022, Macau, China

This article puts forward a methodology for procuring flexible ramping products (FRPs) in the day-ahead market (DAM). The proposed methodology comprises two market passes, the first of which employs a stochastic unit commitment (SUC) model that explicitly evaluates the uncertainty and the intra-hourly and inter-hourly variability of net load so as to minimize the expected total operating cost. The second pass clears the DAM while imposing FRP requirements. The cornerstone of our work is to set the FRP requirements at levels that drive the DAM decisions toward the optimal SUC decisions. Our methodology provides an economic underpinning for the stipulated FRP requirements, and it brings forth DAM awards that reduce the costs toward the expected total cost under SUC, while conforming to the chief DAM design principles. By preemptively considering the dispatch costs before awarding FRPs, it can further avert unexpectedly high costs that could result with the deployment of procured FRPs. We conduct numerical studies and lay out the relative merits of the proposed methodology vis-\`a-vis selected benchmarks based on various evaluation metrics., Comment: 29 pages

Virtualization provides an abstraction layer for the Internet of Things technology to tackle the heterogeneity of the edge networks. It enables the deployment of an application on devices with different architectures to achieve uniformity. This study lays down the fundamentals of a framework for dynamic and event-driven orchestration towards a fully decentralized edge. It provides a blockchain-based delivery platform for containerized applications registered with their resource requirements through a registry on a distributed file system, namely InterPlanetary File System (IPFS). The decentralized resource manager running on the metrics scraped from the host and the virtualization platform, i.e., Docker in our implementation, dynamically optimizes the resources allocated to each container. The framework ensures that variable workloads of a heterogeneous environment can co-exist on multiple edge devices. An event-driven architecture is built over a lightweight messaging protocol, MQTT, capitalizing on the asynchronous and distributed nature of the publish/subscribe pattern to achieve a truly distributed system.

Remote voting has become more critical in recent years, especially after the Covid-19 outbreak. Blockchain technology and its benefits like decentralization, security, and transparency have encouraged remote voting systems to use blockchains. Analysis of existing solutions reveals that anonymity, robustness, and scalability are common problems in blockchain-based election systems. In this work, we propose ElectAnon, a blockchain-based, ranked-choice election protocol focusing on anonymity, robustness, and scalability. ElectAnon achieves anonymity by enabling voters to cast their votes via zero-knowledge proofs anonymously. Robustness is realized by removing the direct control of the authorities in the voting process by using timed-state machines. Results show that ElectAnon is scalable amongst existing works as it reduces the gas consumption up to 89% compared to previous works. The proposed protocol includes a candidate proposal system and swappable tallying libraries. An extension is also proposed to minimize the trust assumption on election authorities. Our code is available on https://github.com/ceyonur/electanon.

For image classification problems, various neural network models are commonly used due to their success in yielding high accuracies. Convolutional Neural Network (CNN) is one of the most frequently used deep learning methods for image classification applications. It may produce extraordinarily accurate results with regard to its complexity. However, the more complex the model is the longer it takes to train. In this paper, an acceleration design that uses the power of FPGA is given for a basic CNN model which consists of one convolutional layer and one fully connected layer for the training phase of the fully connected layer. Nonetheless, inference phase is also accelerated automatically due to the fact that training phase includes inference. In this design, the convolutional layer is calculated by the host computer and the fully connected layer is calculated by an FPGA board. It should be noted that the training of convolutional layer is not taken into account in this design and is left for future research. The results are quite encouraging as this FPGA design tops the performance of some of the state-of-the-art deep learning platforms such as Tensorflow on the host computer approximately 2 times in both training and inference., Comment: Presented at Basarim 2020 conference

We describe a method of low-coherence interferometry based optical profilometry using standard light-emitting diode (LED) illumination and complementary metal-oxide-semiconductor (CMOS) image sensors. A line-field illumination strategy allows for the simultaneous measurement of many points in space. Micron scale accuracy and resolution are achieved and demonstrated using a variety of targets.

Semiconductor colloidal quantum wells (CQWs) make an exciting quasi-2D class of nanocrystals thanks to their unique properties including their highly anisotropic optical transition dipole moment (TDM). Thus, employing a film of CQWs with face-down orientation as an emissive layer (EML) in an electroluminescent device is expected to substantially boost photon outcoupling efficiency. Here, we show all-solution-processed colloidal quantum well light-emitting diodes (CQW-LEDs) using a single all-face-down oriented self-assembled monolayer (SAM) film of CQWs that enables a high level of in-plane (IP) TDMs of 92%. This SAM film significantly enhances the outcoupling efficiency from 22% (of standard randomly-oriented emitters) to 34% (of face-down oriented emitters) and charge injection efficiency. This SAM-CQW-LED architecture enables a record high level of external quantum efficiency of 18.1% for the solution-processed type of CQW-LEDs, putting their efficiency performance on par with the hybrid organic-inorganic evaporation-based CQW-LEDs and all other best solution-processed LEDs. In addition, this architecture provides a high maximum brightness of 19,800 cd/m2 with a long operational lifetime of 247 h at 100 cd/m2 along with saturated deep-red emission (651 nm). These findings indicate the effectiveness of oriented self-assembly of CQWs as electrically-driven emissive layers in improving outcoupling and external quantum efficiencies in the CQW-LEDs.

Interferometric scattering microscopy has been a very promising technology for highly sensitive label-free imaging of a broad spectrum of biological nanoparticles from proteins to viruses in a high-throughput manner. Although it can reveal the specimen's size and shape information, the chemical composition is inaccessible in interferometric measurements. Infrared spectroscopic imaging provides chemical specificity based on inherent chemical bond vibrations of specimens but lacks the ability to image and resolve individual nanoparticles due to long infrared wavelengths. Here, we describe a bond-selective interferometric scattering microscope where the mid-infrared induced photothermal signal is detected by a visible beam in a wide-field common-path interferometry configuration. A thin film layered substrate is utilized to reduce the reflected light and provide a reference field for the interferometric detection of the weakly scattered field. A pulsed mid-IR laser is employed to modulate the interferometric signal. Subsequent demodulation via a virtual lock-in camera offers simultaneous chemical information about tens of micro- or nano-particles. The chemical contrast arises from a minute change in the particle's scattered field in consequence of the vibrational absorption at the target molecule. We characterize the system with sub-wavelength polymer beads and highlight biological applications by chemically imaging several microorganisms including Staphylococcus aureus, Escherichia coli, and Candida albicans. A theoretical framework is established to extend bond-selective interferometric scattering microscopy to a broad range of biological micro- and nano-particles.

Mid-infrared photothermal (MIP) microscopy has been a promising label-free chemical imaging technique for functional characterization of specimens owing to its enhanced spatial resolution and high specificity. Recently developed wide-field MIP imaging modalities have drastically improved speed and enabled high-throughput imaging of micron-scale subjects. However, the weakly scattered signal from sub-wavelength particles becomes indistinguishable from the shot-noise as a consequence of the strong background light, leading to limited sensitivity. Here, we demonstrate background-suppressed chemical fingerprinting at a single nanoparticle level by selectively attenuating the reflected light through pupil engineering in the collection path. Our technique provides over three orders of magnitude background suppression by quasi-darkfield illumination in epi-configuration without sacrificing lateral resolution. We demonstrate 6-fold signal-to-background noise ratio improvement, allowing for simultaneous detection and discrimination of hundreds of nanoparticles across a field of view of 70 um x 70 um. A comprehensive theoretical framework for photothermal image formation is provided and experimentally validated with 300 and 500~nm PMMA beads. The versatility and utility of our technique are demonstrated via hyperspectral dark-field MIP imaging of S. aureus and E. coli bacteria., Comment: 33 pages, 7 figures, journal article

Applications of reinforcement learning (RL) are popular in autonomous driving tasks. That being said, tuning the performance of an RL agent and guaranteeing the generalization performance across variety of different driving scenarios is still largely an open problem. In particular, getting good performance on complex road and weather conditions require exhaustive tuning and computation time. Curriculum RL, which focuses on solving simpler automation tasks in order to transfer knowledge to complex tasks, is attracting attention in RL community. The main contribution of this paper is a systematic study for investigating the value of curriculum reinforcement learning in autonomous driving applications. For this purpose, we setup several different driving scenarios in a realistic driving simulator, with varying road complexity and weather conditions. Next, we train and evaluate performance of RL agents on different sequences of task combinations and curricula. Results show that curriculum RL can yield significant gains in complex driving tasks, both in terms of driving performance and sample complexity. Results also demonstrate that different curricula might enable different benefits, which hints future research directions for automated curriculum training., Comment: 6 pages, IV2021 Workshop