Search

Adaptive learning is necessary for non-stationary environments where the learning machine needs to forget past data distribution. Efficient algorithms require a compact model update to not grow in computational burden with the incoming data and with the lowest possible computational cost for online parameter updating. Existing solutions only partially cover these needs. Here, we propose the first adaptive sparse Gaussian Process (GP) able to address all these issues. We first reformulate a variational sparse GP algorithm to make it adaptive through a forgetting factor. Next, to make the model inference as simple as possible, we propose updating a single inducing point of the sparse GP model together with the remaining model parameters every time a new sample arrives. As a result, the algorithm presents a fast convergence of the inference process, which allows an efficient model update (with a single inference iteration) even in highly non-stationary environments. Experimental results demonstrate the capabilities of the proposed algorithm and its good performance in modeling the predictive posterior in mean and confidence interval estimation compared to state-of-the-art approaches.

Frequency reuse in cells is one of the strategies that LTE uses to maximize the spectrum efficiency. However, it leads to an interference among the cells, especially at the cell edges where the probability for a cell-edge user to be scheduled on a resource block that is being transmitted by the neighbouring cell is high; consequently, the interference is high. In-order to mitigate Inter-Cell Interference, Inter-Cell Interference Coordination was proposed by the Third Generation Partnership Project 3GPP standards for the LTE network, and later on, the enhanced Inter-Cell Interference Coordination was proposed for the LTE-Advanced Heterogeneous Network deployments. Another technique that was proposed by the 3GPP for the LTE-Advanced network that can reduce the Inter-Cell Interference and improve the cell-average and cell-edge user throughput is the Coordinated Multi-Point transmission and reception technique, in which it can increase the cell-average and cell-edge user throughput in both the uplink and downlink transmission by joint scheduling and data processing in multiple cells. Another technique that was proposed by the 3GPP for the LTE-Advanced network that can improve the LTE network coverage in difficult conditions is to deploy Relay Nodes and Back-Hauls. In this paper, we survey the Radio Resource Management of these techniques that are used with LTE-Advanced. We start by explaining the concepts of these techniques. Then, by summarizing the radio resource management approaches that were proposed in the literature for these techniques. And finally, we provide some concluding remarks in the last section.

The number of LTE users and their applications has increased significantly in the last decade, which increased the demand on the mobile network. LTE-Advanced comes with many features that can support this increasing demand. LTE-Advanced supports Heterogeneous Networks deployment, in which it consists of a mix of macro-cells, remote radio heads, and low power nodes such as Pico-cells, and Femto-cells. Embedding this mix of base-stations in a macro-cellular network allows for achieving significant gains in coverage, throughput and system capacity compared to the use of macro-cells only. These base-stations can operate on the same wireless channel as the macro-cellular network, which will provide higher spatial reuse via cell splitting. Also, it allows network operators to support higher data traffic by offloading it to smaller cells, such as Femto-cells. Hence, it enables network operators to provide their growing number of users with the required Quality of Service that meets with their service demands. In-order for the network operators to make the best out of the heterogeneous LTE-Advanced network, they need to use QoS-based packet scheduling algorithms that can efficiently manage the spectrum resources in the heterogeneous deployment. In this paper, we survey Quality of Service based packet scheduling algorithms that were proposed in the literature for the use of packet scheduling in Heterogeneous LTE-Advanced Networks. We start by explaining the concepts of QoS in LTE, heterogeneous LTE-Advanced networks, and how traffic is classified within a packet scheduling architecture for heterogeneous LTE-Advanced networks. Then, by summarising the proposed QoS-based packet scheduling algorithms in the literature for heterogeneous LTE-Advanced Networks, and for Femtocells LTE-Advanced Networks. And finally, we provide some concluding remarks in the last section.

This survey paper provides a detailed explanation of Long Term Evolution cellular network packet scheduling algorithms in both downlink and uplink directions. It starts by explaining the difference between Orthogonal Frequency Division Multiple Access that is used in downlink transmission, and Single Carrier Frequency Division Multiple Access is used in uplink. Then, it explains the difference between the LTE scheduling process in the donwlink and uplink through explaining the interaction between users and the scheduler. Then, it explains the most commonly used downlink and uplink scheduling algorithms through analyzing their formulas, characteristics, most suitable conditions for them to work in, and the main differences among them. This explanation covers the Max Carrier-to-Interference, Round Robin, Proportional Fair, Earliest Deadline First, Modified EDF-PF, Modified-Largest Weighted Delay First, Exponential Proportional Fairness, Token Queues Mechanism, Packet Loss Ratio, Quality Guaranteed, Opportunistic Packet Loss Fair, Low Complexity, LC-Delay, PF-Delay, Maximum Throughput, First Maximum Expansion, and Adaptive Resource Allocation Based Packet Scheduling. Lastly, it provides some concluding remarks.

In this paper, we propose, implement, and test two novel downlink LTE scheduling algorithms. The implementation and testing of these algorithms were in Matlab, and they are based on the use of Reinforcement Learning, more specifically, the Qlearning technique for scheduling two types of users. The first algorithm is called a Collaborative scheduling algorithm, and the second algorithm is called a Competitive scheduling algorithm. The first type of the scheduled users is the Primary Users, and they are the licensed subscribers that pay for their service. The second type of the scheduled users is the Secondary Users, and they could be unlicensed subscribers that dont pay for their service, device to device communications, or sensors. Each user whether it is a primary or secondary is considered as an agent. In the Collaborative scheduling algorithm, the primary user agents will collaborate in order to make a joint scheduling decision about allocating the resource blocks to each one of them, then the secondary user agents will compete among themselves to use the remaining resource blocks. In the Competitive scheduling algorithm, the primary user agents will compete among themselves over the available resources, then the secondary user agents will compete among themselves over the remaining resources. Experimental results show that both scheduling algorithms converged to almost ninety percent utilization of the spectrum, and provided fair shares of the spectrum among users.

The uncertainty of the energy generated by photovoltaic systems incurs an additional cost for a guaranteed, reliable supply of energy (i.e., energy storage). This investigation aims to decrease the additional cost by introducing probabilistic multi-task intra-hour solar forecasting (feasible in real time applications) to increase the penetration of photovoltaic systems in power grids. The direction of moving clouds is estimated in consecutive sequences of sky images by extracting features of cloud dynamics with the objective of forecasting the global solar irradiance that reaches photovoltaic systems. The sky images are acquired using a low-cost infrared sky imager mounted on a solar tracker. The solar forecasting algorithm is based on kernel learning methods, and uses the clear sky index as predictor and features extracted from clouds as feature vectors. The proposed solar forecasting algorithm achieved 16.45\% forecasting skill 8 minutes ahead with a resolution of 15 seconds. In contrast, previous work reached 15.4\% forecasting skill with the resolution of 1 minute. Therefore, this solar forecasting algorithm increases the performances with respect to the state-of-the-art, providing grid operators with the capability of managing the inherent uncertainties of power grids with a high penetration of photovoltaic systems.

This work investigates application of different machine learning based prediction methodologies to estimate the performance of silicon based textured cells. Concept of confidence bound regions is introduced and advantages of this concept are discussed in detail. Results show that reflection profiles and depth dependent optical generation profiles can be accurately estimated using Gaussian processes with exact knowledge of uncertainty in the prediction values.It is also shown that cell design parameters can be estimated for a desired performance metric.

Horizontal atmospheric wind shear causes wind velocity fields to have different directions and speeds. In images of clouds acquired using ground-based sky imagers, clouds may be moving in different wind layers. To increase the performance of an intra-hour global solar irradiance forecasting algorithm, it is important to detect multiple layers of clouds. The information provided by a solar forecasting algorithm is necessary to optimize and schedule the solar generation resources and storage devices in a smart grid. This investigation studies the performance of unsupervised learning techniques when detecting the number of cloud layers in infrared sky images. The images are acquired using an innovative infrared sky imager mounted on a solar tracker. Different mixture models are used to infer the distribution of the cloud features. The optimal decision criterion to find the number of clusters in the mixture models is analyzed and compared between different Bayesian metrics and a sequential hidden Markov model. The motion vectors are computed using a weighted implementation of the Lucas-Kanade algorithm. The correlations between the cloud velocity vectors and temperatures are analyzed to find the method that leads to the most accurate results. We have found that the sequential hidden Markov model outperformed the detection accuracy of the Bayesian metrics.

Moving clouds affect the global solar irradiance that reaches the surface of the Earth. As a consequence, the amount of resources available to meet the energy demand in a smart grid powered using Photovoltaic (PV) systems depends on the shadows projected by passing clouds. This research introduces an algorithm for tracking clouds to predict Sun occlusion. Using thermal images of clouds, the algorithm is capable of estimating multiple wind velocity fields with different altitudes, velocity magnitudes and directions., Comment: arXiv admin note: text overlap with arXiv:2012.02861

Sky imaging systems use lenses to acquire images concentrating light beams in a sensor. The light beams received by the sky imager have an elevation angle with respect to the device normal. Thus, the pixels in the image contain information from different areas of the sky within the imaging system field of view. The area of the field of view contained in the pixels increases as the elevation angle of the incident light beams decreases. When the sky imager is mounted on a solar tracker, the light beam's angle of incidence in a pixel varies over time. This investigation formulates and compares two geospatial reprojections that transform the original euclidean frame of the imager plane to the geospatial atmosphere cross-section where the sky imager field of view intersects the cloud layer. One assumes that an object (i.e., cloud) moving in the troposphere is sufficiently far so the Earth's surface is approximated \emph{flat}. The other transformation takes into account the curvature of the Earth in the portion of the atmosphere (i.e., voxel) that is recorded. The results show that the differences between the dimensions calculated by both geospatial transformations are in the order of magnitude of kilometers when the Sun's elevation angle is below $30^\circ$.

The increasing number of Photovoltaic (PV) systems connected to the power grid are vulnerable to the projection of shadows from moving clouds. Global Solar Irradiance (GSI) forecasting allows smart grids to optimize the energy dispatch, preventing energy shortages caused by occlusion of the sun. This investigation compares the performances of machine learning algorithms (not requiring labelled images for training) for real-time segmentation of clouds in images acquired using a ground-based infrared sky imager. Real-time segmentation is utilized to extract cloud features using only the pixels in which clouds are detected.

The energy available in Micro Grid (MG) that is powered by solar energy is tightly related to the weather conditions in the moment of generation. Very short-term forecast of solar irradiance provides the MG with the capability of automatically controlling the dispatch of energy. We propose to achieve this using a data acquisition systems (DAQ) that simultaneously records sky imaging and Global Solar Irradiance (GSI) measurements, with the objective of extracting features from clouds and use them to forecast the power produced by a Photovoltaic (PV) system. The DAQ system is nicknamed as the \emph{Girasol Machine} (Girasol means Sunflower in Spanish). The sky imaging system consists of a longwave infrared (IR) camera and a visible (VI) light camera with a fisheye lens attached to it. The cameras are installed inside a weatherproof enclosure that it is mounted on an outdoor tracker. The tracker updates its pan an tilt every second using a solar position algorithm to maintain the Sun in the center of the IR and VI images. A pyranometer is situated on a horizontal support next to the DAQ system to measure GSI. The dataset, composed of IR images, VI images, GSI measurements, and the Sun's positions, has been tagged with timestamps., Comment: arXiv admin note: substantial text overlap with arXiv:2011.12401

Photovoltaic systems are sensitive to cloud shadow projection, which needs to be forecasted to reduce the noise impacting the intra-hour forecast of global solar irradiance. We present a comparison between different kernel discriminative models for cloud detection. The models are solved in the primal formulation to make them feasible in real-time applications. The performances are compared using the j-statistic. The infrared cloud images have been preprocessed to remove debris, which increases the performance of the analyzed methods. The use of neighboring features of the pixels also leads to a performance improvement. Discriminative models solved in the primal yield a dramatically lower computing time along with high performance in the segmentation.

The projection of shadows from moving clouds in the troposphere impacts energy generation in power grids using photovoltaic systems. This investigation proposes an efficient method of data processing for the statistical quantification of cloud features using infrared images and global solar irradiance measurements. The infrared images and the global solar irradiance measurements are acquired using a sky imager equipped with a commercial low-cost long-wave infrared radiometric camera and a pyranometer. The enclosure of the infrared camera is mounted on a solar tracker so that the Sun stays in the center of images throughout the day. We explain how to remove cyclostationary biases in global solar irradiance measurements. Seasonal trends are removed from the global solar irradiance time series, using the theoretical global solar irradiance to obtain the clear sky index time series. We introduce an atmospheric model to remove the effect of atmospheric scattering and the effect of the Sun's direct irradiance from infrared images. Scattering is produced by water spots and dust particles on the germanium lens of the camera enclosure. We explain how to remove the scattering effect produced by the germanium lens attached to the data acquisition system enclosure window of the infrared camera. An atmospheric condition model classifies the sky conditions in four different categories: clear sky, cumulus, stratus and nimbus. When an infrared image is classified in the category of clear sky, it is used to model the scattering effect produced by the germanium lens., Comment: Incorrect formulation

The increasing penetration of photovoltaic systems in the power grid makes it vulnerable to cloud shadow projection. Real-time cloud segmentation in ground-based infrared images is important to reduce the noise in intra-hour global solar irradiance forecasting. We present a comparison between discriminative and generative models for cloud segmentation. The performances of supervised and unsupervised learning methods in cloud segmentation are evaluated. The discriminative models are solved in the primal formulation to make them feasible in real-time applications. The performances are compared using the j-statistic. Infrared image preprocessing to remove stationary artifacts increases the overall performance in the analyzed methods. The inclusion of features from neighboring pixels in the feature vectors leads to a performance improvement in some of the cases. Markov Random Fields achieve the best performance in both unsupervised and supervised generative models. Discriminative models solved in the primal yield a dramatically lower computing time along with high performance in the segmentation. Generative and discriminative models are comparable when preprocessing is applied to the infrared images.

The energy available in a solar energy powered grid is uncertain due to the weather conditions at the time of generation. Forecasting global solar irradiance could address this problem by providing the power grid with the capability of scheduling the storage and dispatch of energy. The occlusion of the Sun by clouds is the main cause of instabilities in the generation of solar energy. This investigation proposes a method to visualize the wind velocity field in sequences of longwave infrared images of clouds when there are multiple wind velocity fields in an image. This method can be used to forecast the occlusion of the Sun by clouds, providing stability in the generation of solar energy. Unsupervised learning is implemented to infer the distribution of the clouds' velocity vectors and heights in multiple wind velocity fields in an infrared image. A multi-output weighted support vector machine with flow constraints is used to extrapolate the wind velocity fields to the entire frame, visualizing the path of the clouds. The proposed method is capable of approximating the wind velocity field in a small air parcel using the velocity vectors and physical features of clouds extracted from infrared images. Assuming that the streamlines are pathlines, the visualization of the wind velocity field can be used for forecasting cloud occlusions of the Sun. This is of importance when considering ways of increasing the stability of solar energy generation.

The energy available in Micro Grid (MG) that is powered by solar energy is tightly related to the weather conditions in the moment of generation. Very short-term forecast of solar irradiance provides the MG with the capability of automatically controlling the dispatch of energy. To achieve this, we propose a method for statistical quantification of cloud features extracted from long-ware infrared (IR) images to forecast the Clear Sky Index (CSI). The images are obtained using a data acquisition system (DAQ) mounted on a solar tracker. We explain how to remove cyclostationary bias in the data caused by the devices in the own DAQ. We investigate a method to obtain the CSI, after the detrending of Global Horizontal Irradiance (GHI) measurements. We propose a method to fusion multiple exposures of circumsolar visible (VI) light images. We implement a method for extracting physical features using radiometric measurements of the IR camera. We introduce a model to remove from IR images both the effect of the atmosphere scatter radiation, and the effect of the Sun direct radiation. We explain how to model of diffuse radiation of the IR camera window, which is produce by water spots and dust particles stack to the germanium lens of the DAQ enclosure. The frames, that were used to model the camera window, are selected using an atmospheric condition model. This model classifies the sky four different categories: clear, cumulus, stratus, and nimbus. We introduce a geometric transformation of the size of the pixels to their actual dimension in a plane of the atmosphere which is at a given height. This transformation is performed according to the elevation angle of the Sun and field of view (FOV) of the camera. We compare the error between the transformation and anapproximation of transformation.

Live fire creates a dynamic, rapidly changing environment that presents a worthy challenge for deep learning and artificial intelligence methodologies to assist firefighters with scene comprehension in maintaining their situational awareness, tracking and relay of important features necessary for key decisions as they tackle these catastrophic events. We propose a deep Q-learning based agent who is immune to stress induced disorientation and anxiety and thus able to make clear decisions for navigation based on the observed and stored facts in live fire environments. As a proof of concept, we imitate structural fire in a gaming engine called Unreal Engine which enables the interaction of the agent with the environment. The agent is trained with a deep Q-learning algorithm based on a set of rewards and penalties as per its actions on the environment. We exploit experience replay to accelerate the learning process and augment the learning of the agent with human-derived experiences. The agent trained under this deep Q-learning approach outperforms agents trained through alternative path planning systems and demonstrates this methodology as a promising foundation on which to build a path planning navigation assistant capable of safely guiding fire fighters through live fire environments., Comment: Accepted to ICAART2021

Firefighting is a dynamic activity, in which numerous operations occur simultaneously. Maintaining situational awareness (i.e., knowledge of current conditions and activities at the scene) is critical to the accurate decision-making necessary for the safe and successful navigation of a fire environment by firefighters. Conversely, the disorientation caused by hazards such as smoke and extreme heat can lead to injury or even fatality. This research implements recent advancements in technology such as deep learning, point cloud and thermal imaging, and augmented reality platforms to improve a firefighter's situational awareness and scene navigation through improved interpretation of that scene. We have designed and built a prototype embedded system that can leverage data streamed from cameras built into a firefighter's personal protective equipment (PPE) to capture thermal, RGB color, and depth imagery and then deploy already developed deep learning models to analyze the input data in real time. The embedded system analyzes and returns the processed images via wireless streaming, where they can be viewed remotely and relayed back to the firefighter using an augmented reality platform that visualizes the results of the analyzed inputs and draws the firefighter's attention to objects of interest, such as doors and windows otherwise invisible through smoke and flames., Comment: Accepted to ICMLA Special Session on Deep Learning

Situational awareness and Indoor location tracking for firefighters is one of the tasks with paramount importance in search and rescue operations. For Indoor Positioning systems (IPS), GPS is not the best possible solution. There are few other techniques like dead reckoning, Wifi and bluetooth based triangulation, Structure from Motion (SFM) based scene reconstruction for Indoor positioning system. However due to high temperatures, the rapidly changing environment of fires, and low parallax in the thermal images, these techniques are not suitable for relaying the necessary information in a fire fighting environment needed to increase situational awareness in real time. In fire fighting environments, thermal imaging cameras are used due to smoke and low visibility hence obtaining relative orientation from the vanishing point estimation is very difficult. The following technique that is the content of this research implements a novel optical flow based video compass for orientation estimation and fused IMU data based activity recognition for IPS. This technique helps first responders to go into unprepared, unknown environments and still maintain situational awareness like the orientation and, position of the victim fire fighters.

Multitask Gaussian processes (MTGP) are the Gaussian process (GP) framework's solution for multioutput regression problems in which the $T$ elements of the regressors cannot be considered conditionally independent given the observations. Standard MTGP models assume that there exist both a multitask covariance matrix as a function of an intertask matrix, and a noise covariance matrix. These matrices need to be approximated by a low rank simplification of order $P$ in order to reduce the number of parameters to be learnt from $T^2$ to $TP$. Here we introduce a novel approach that simplifies the multitask learning by reducing it to a set of conditioned univariate GPs without the need for any low rank approximations, therefore completely eliminating the requirement to select an adequate value for hyperparameter $P$. At the same time, by extending this approach with both a hierarchical and an approximate model, the proposed extensions are capable of recovering the multitask covariance and noise matrices after learning only $2T$ parameters, avoiding the validation of any model hyperparameter and reducing the overall complexity of the model as well as the risk of overfitting. Experimental results over synthetic and real problems confirm the advantages of this inference approach in its ability to accurately recover the original noise and signal matrices, as well as the achieved performance improvement in comparison to other state of art MTGP approaches. We have also integrated the model with standard GP toolboxes, showing that it is computationally competitive with state of the art options.

The recent advances in deep learning indicate significant progress in the field of single image super-resolution. With the advent of these techniques, high-resolution image with high peak signal to noise ratio (PSNR) and excellent perceptual quality can be reconstructed. The major challenges associated with existing deep convolutional neural networks are their computational complexity and time; the increasing depth of the networks, often result in high space complexity. To alleviate these issues, we developed an innovative shallow residual feature representative network (SRFRN) that uses a bicubic interpolated low-resolution image as input and residual representative units (RFR) which include serially stacked residual non-linear convolutions. Furthermore, the reconstruction of the high-resolution image is done by combining the output of the RFR units and the residual output from the bicubic interpolated LR image. Finally, multiple experiments have been performed on the benchmark datasets and the proposed model illustrates superior performance for higher scales. Besides, this model also exhibits faster execution time compared to all the existing approaches.

Koopman Mode Decomposition (KMD) is a technique of nonlinear time-series analysis that originates from point spectrum of the Koopman operator defined for an underlying nonlinear dynamical system. We present a numerical algorithm of KMD based on Gaussian process regression that is capable of handling noisy finite-time data. The algorithm is applied to short-term swing dynamics of a multi-machine power grid in order to estimate oscillatory modes embedded in the dynamics, and thereby the effectiveness of the algorithm is evaluated., Comment: 7 pages, 6 figures, 2019 Japan Joint Automatic Control Conference

Superconducting Radio-Frequency (SRF) cavities operating in continuous wave (CW) mode and with low beam loading are devices characterized by a high loaded quality factor, in the order of 10^7, and narrow bandwidth, in the order of 10 Hz. The Low Level RF (LLRF) and resonance control systems of SRF cavities become a fundamental component of the entire system operation and in general has very tight stability requirements on the amplitude, phase and resonance frequency of the cavity. Microphonics plays an important role in cavity detuning, which results in more power needed to achieve the desired gradient. Active control approaches to reduce detuning in SRF cavities using piezoelectric actuators have shown promising results. Furthermore, Machine Learning (ML) techniques have also shown important capabilities to improve existing PID controllers. Specifically, Neural Networks (NN) can be used to find optimal PID gains and improve performance of traditional control systems. In this research, we develop new approaches to improve existing LLRF and resonance control systems with ML as a tool to find the optimal gains. We investigate these approaches using the LLRF control system intended for LCLS-II., Comment: Talk and Poster presented at LLRF Workshop 2019 (LLRF2019, arXiv:1909.06754)

Intelligent detection and processing capabilities can be instrumental to improving the safety, efficiency, and successful completion of rescue missions conducted by firefighters in emergency first response settings. The objective of this research is to create an automated system that is capable of real-time, intelligent object detection and recognition and facilitates the improved situational awareness of firefighters during an emergency response. We have explored state of the art machine/deep learning techniques to achieve this objective. The goal for this work is to enhance the situational awareness of firefighters by effectively exploiting the information gathered from infrared cameras carried by firefighters. To accomplish this, we use a trained deep Convolutional Neural Network (CNN) system to classify and identify objects of interest from thermal imagery in real time. In the midst of those critical circumstances created by structure fire, this system is able to accurately inform the decision making process of firefighters with real-time up-to-date scene information by extracting, processing, and analyzing crucial information. With the new information produced by the framework, firefighters are able to make more informed inferences about the circumstances for their safe navigation through such hazardous and potentially catastrophic environments.