Computational Core Research Articles

The scalability analysis of a parallel tree search algorithm

Increasing the number of computational cores is a primary way of achieving the high performance of contemporary supercomputers. However, developing parallel applications capable to harness the enormous amount of cores is a challenging task. It is very important to understand the principle limitations of the scalability of parallel applications imposed by the algorithm’s structure. The tree search addressed in this paper has an irregular structure unknown prior to computations. That is why such algorithms are challenging for parallel implementation especially on distributed memory systems. In this paper, we propose a parallel tree search algorithm aimed at distributed memory parallel computers. For this parallel algorithm, we analyze its scalability and show that it is close to the theoretical maximum.

A Simple Control Scheme for Improving Reactive Power Sharing in Islanded Microgrid

In this article, a simple control scheme based on an adaptive $Q-V$ droop coefficient is proposed to improve the reactive power sharing amongst sources in an islanded microgrid. All the sources present in the network independently calculate and apply an adaptive $Q-V$ droop coefficient after every step change in load, thereby achieving system-wide improved reactive power sharing. The core computation of the proposed method is based on the transient energy delivered by the respective inverter after every load change. The proposed control does not require communication; neither depends on the knowledge of feeder impedances nor requires measurements from distant nodes. Also, it can work satisfactorily irrespective of the location of the load change. Further, the strategy is also immune to any load change happening during the compensation process. A significant improvement in reactive power sharing has been achieved by using the proposed scheme without compromising the voltage regulation, and without causing significant power transients. The effectiveness of the proposed scheme has been validated by performing detailed simulation studies. Its viability is confirmed by carrying out experimental studies on a laboratory prototype.

Deep Learning for Real-Time 3D Multi-Object Detection, Localisation, and Tracking: Application to Smart Mobility.

In core computer vision tasks, we have witnessed significant advances in object detection, localisation and tracking. However, there are currently no methods to detect, localize and track objects in road environments, and taking into account real-time constraints. In this paper, our objective is to develop a deep learning multi object detection and tracking technique applied to road smart mobility. Firstly, we propose an effective detector-based on YOLOv3 which we adapt to our context. Subsequently, to localize successfully the detected objects, we put forward an adaptive method aiming to extract 3D information, i.e., depth maps. To do so, a comparative study is carried out taking into account two approaches: Monodepth2 for monocular vision and MADNEt for stereoscopic vision. These approaches are then evaluated over datasets containing depth information in order to discern the best solution that performs better in real-time conditions. Object tracking is necessary in order to mitigate the risks of collisions. Unlike traditional tracking approaches which require target initialization beforehand, our approach consists of using information from object detection and distance estimation to initialize targets and to track them later. Expressly, we propose here to improve SORT approach for 3D object tracking. We introduce an extended Kalman filter to better estimate the position of objects. Extensive experiments carried out on KITTI dataset prove that our proposal outperforms state-of-the-art approches.

EplusLauncher: An API to Perform Complex EnergyPlus Simulations in MATLAB® and C#

There is a growing concern about how to mitigate climate change, in which the production and use of energy has a great impact as one of the largest sources of global greenhouse gases (GHG). Buildings are responsible for a large percentage of these emissions. Therefore, there has been an increase in research in this area, in order to reduce their consumption and increase their efficiency. One of the major simulation programs used in optimization research is EnergyPlus. The purpose of this software is the complete energy simulation of a building, although it lacks tools to analyze its results and, above all, to manage and edit its simulations. For this reason, we developed an application programming interface (API) that serves to merge two areas which are highly demanded by researchers: energy building simulation (using EnergyPlus) and tools for the management and design of research experiments (in this case, MATLAB®). The developed API allows the user to perform complex simulations using EnergyPlus in a simple way, as it allows the editing of each simulation and the analysis of the simulation results through MATLAB®. In addition, it enables the user to simultaneously run multiple simulations, using either all computer core processors or a selection of them (i.e., allowing parallel computing), reducing the simulation time. The API was developed in the C# language, such that it can be used with any software that can import . N E T libraries.

WInc3D: A novel framework for turbulence‐resolving simulations of wind farm wake interactions

AbstractA fast and efficient turbulence‐resolving computational framework, dubbed as WInc3D (Wind Incompressible 3‐Dimensional solver), is presented and validated in this paper. WInc3D offers a unified, highly scalable, high‐fidelity framework for the study of the flow structures and turbulence of wind farm wakes and their impact on the individual turbines' power and loads. Its unique properties lie on the use of higher‐order numerical schemes with “spectral‐like” accuracy, a highly efficient parallelisation strategy which allows the code to scale up to O(104) computing processors and software compactness (use of only native solvers/models) with virtually no dependence to external libraries. The work presents an overview of the current modelling capabilities along with model validation. The presented applications demonstrate the ability of WInc3D to be used for testing farm‐level optimal control strategies using turbine wakes under yawed conditions. Examples are provided for two turbines operating in‐line as well as a small array of 16 turbines operating under “Greedy” and “Co‐operative” yaw angle settings. These large‐scale simulations were performed with up to 8192 computational cores for under 24 hours, for a computational domain discretised with O(109) mesh nodes.

A Magnetometer-Only Attitude Determination Strategy for Small Satellites: Design of the Algorithm and Hardware-in-the-Loop Testing

Attitude determination represents a fundamental task for spacecraft. Achieving this task on small satellites, and nanosatellites in particular, is further challenging, because the limited power and computational resources available on-board, together with the low development budget, set strict constraints on the selection of the sensors and the complexity of the algorithms. Attitude determination is obtained here from the only measurements of a three-axis magnetometer and a model of the Geomagnetic field, stored on the on-board computer. First, the angular rates are estimated and processed using a second-order low-pass Butterworth filter, then they are used as an input, along with Geomagnetic field data, to estimate the attitude matrix using an unsymmetrical TRIAD. The computational efficiency is enhanced by arranging complex matrix operations into a form of the Faddeev algorithm, which is implemented using systolic array architecture on the FPGA core of a CubeSat on-board computer. The performance and the robustness of the algorithm are evaluated by means of numerical analyses in MATLAB Simulink, showing pointing and angular rate accuracy below 10° and 0.2°/s. The algorithm implemented on FPGA is verified by Hardware-in-the-loop simulation, confirming the results from numerical analyses and efficiency.

MultiBUGS: A Parallel Implementation of the BUGS Modelling Framework for Faster Bayesian Inference.

MultiBUGS is a new version of the general-purpose Bayesian modelling software BUGS that implements a generic algorithm for parallelising Markov chain Monte Carlo (MCMC) algorithms to speed up posterior inference of Bayesian models. The algorithm parallelises evaluation of the product-form likelihoods formed when a parameter has many children in the directed acyclic graph (DAG) representation; and parallelises sampling of conditionally-independent sets of parameters. A heuristic algorithm is used to decide which approach to use for each parameter and to apportion computation across computational cores. This enables MultiBUGS to automatically parallelise the broad range of statistical models that can be fitted using BUGS-language software, making the dramatic speed-ups of modern multi-core computing accessible to applied statisticians, without requiring any experience of parallel programming. We demonstrate the use of MultiBUGS on simulated data designed to mimic a hierarchical e-health linked-data study of methadone prescriptions including 425,112 observations and 20,426 random effects. Posterior inference for the e-health model takes several hours in existing software, but MultiBUGS can perform inference in only 28 minutes using 48 computational cores.

The Concept of Information. Part 2. The Functions of a Scientific Theory

Any scientific theory should effectively perform a variety of information processing functions, regardless of the specifics of knowledge and the field of formalization. To achieve the global aim − maximum assistance in solving each of the infinite number of problems of the evolution of society − the information computer should be able to simulate all the functions of scientific theories, which are combined into a conceptually single system of knowledge processing, and due to this the information computer should provide the possibility to use knowledge directly in an explicit form for solving current and new problems. The meaning of communicative, descriptive-prescriptive, and argumentative (deduction, induction, explanation, prediction, understanding) functions of scientific theories is clarified. The models of synthesis of various components of the structure of scientific theories are described. The objects of real-world and the communicative functions of analysis and generation of language messages are enough for the organization of arbitrarily complex dialogue between a subject and the real world. The descriptive-prescriptive functions are the cornerstone for cognition and transformation of reality. All facts of reality are descriptive, whereas the standards, samples, ideals and norms are estimates. The argumentation is a process of reasoning during which the arguments, opinions, and conclusions are generated, that justify the assessment of truth or falsity of a statement or a theory. The self-description of the metalanguage of normal forms of knowledge is presented. The structure of the core of the information computer, which is based on this metalanguage, is presented, comprised of the following: the metaprogram compiler, two interpreters (the metaprogram interpreter and the terminal program interpreter), a glossary of terms and two arrays of data (input array and output array). The process of functioning of a knowledge processing system based on an information computer is described. The information on the implementation of the layout of the information computer in the form of a personal computer supplemented with the knowledge base interpreter made on a prototyping board of Actel corporation using minimum required hardware and software support is given. Giving users the metalanguage for describing application problems and the improved technology of formulating and solving problems, which realize all the features of a scientific theory, will provide extremely high growth of "soft" product reliability and programming efficiency.

A memory-based task scheduling algorithm for grid computing based on heterogeneous platform and homogeneous tasks

Grid computing is a new computing mode in recent years, which focuses on parallel infrastructure and its comprehensive application ability to network computers and distributed processors. Grid computing has been fully applied in the field of modern information technology and computer. Task scheduling is the core of grid computing. The quality of task scheduling algorithm directly affects the response time of the whole computing system. For heterogeneous tasks on heterogeneous platforms, this paper proposes a task scheduling algorithm with memory function, and introduces the distributed particle swarm optimisation algorithm into this algorithm, which realises the combination of resource processing tasks in grid computing and the behaviour characteristics of intelligent groups, so as to better realise the dynamic and scalable scheduling of heterogeneous tasks on heterogeneous platforms to adapt to grid environment sex. Finally, the grid simulation software GridSim is used to simulate the algorithm proposed in this paper. At the same time, it is compared with the state stochastic scheduling algorithm. Experimental results show that the proposed algorithm has obvious advantages in scheduling quality in grid environment.

Convergence of Edge Computing and Deep Learning: A Comprehensive Survey

Ubiquitous sensors and smart devices from factories and communities are generating massive amounts of data, and ever-increasing computing power is driving the core of computation and services from the cloud to the edge of the network. As an important enabler broadly changing people's lives, from face recognition to ambitious smart factories and cities, developments of artificial intelligence (especially deep learning, DL) based applications and services are thriving. However, due to efficiency and latency issues, the current cloud computing service architecture hinders the vision of "providing artificial intelligence for every person and every organization at everywhere". Thus, unleashing DL services using resources at the network edge near the data sources has emerged as a desirable solution. Therefore, edge intelligence, aiming to facilitate the deployment of DL services by edge computing, has received significant attention. In addition, DL, as the representative technique of artificial intelligence, can be integrated into edge computing frameworks to build intelligent edge for dynamic, adaptive edge maintenance and management. With regard to mutually beneficial edge intelligence and intelligent edge, this paper introduces and discusses: 1) the application scenarios of both; 2) the practical implementation methods and enabling technologies, namely DL training and inference in the customized edge computing framework; 3) challenges and future trends of more pervasive and fine-grained intelligence. We believe that by consolidating information scattered across the communication, networking, and DL areas, this survey can help readers to understand the connections between enabling technologies while promoting further discussions on the fusion of edge intelligence and intelligent edge, i.e., Edge DL.

Evolution of the CloudVeneto.it private cloud to support research and innovation

CloudVeneto.it was initially funded and deployed by INFN in 2014 for serving the computational and storage demands of INFN research projects mainly related to HEP and Nuclear Physics. It is an OpenStack-based scientific cloud with resources spread across two different sites connected with a high speed optical link: INFN Padova Unit and the INFN Legnaro National Laboratories. The infrastructure has grown throughout the years with additional funds from ten University of Padova departments, and nowadays supports a broader range of scientific and engineering disciplines. Its hardware resources provide around 2500 computational cores and 360 TB of storage to about 250 users working for more than 70 projects. In the last months we enhanced the cloud platform in two ways: 1) by integrating a number of heterogeneous GPU cards to address the special needs of user communities whose computations involve machine learning training; 2) by enabling the users to simply deploy on-demand Kubernetes clusters for Big Data Analytics applications taking advantage of the operator framework. In particular, the Kubernetes operators for Apache Kafka and Spark platforms were integrated to address real-time data ingestion and streaming processing on the cloud. This article describes the technical details of these two solutions and their integration with the cloud infrastructure.

A study on the design methodology of TAC3 for edge computing.

The following scenarios, such as complex application requirements, ZB (Zettabyte) order of magnitude of network data, and tens of billions of connected devices, pose serious challenges to the capabilities and security of the three pillars of ICT: Computing, network, and storage. Edge computing came into being. Following the design methodology of "description-synthesis-simulation-optimization", TAC3 (Tile-Architecture Cluster Computing Core) was proposed as the lightweight accelerated ECN (Edge Computing Node). ECN with a Tile-Architecture be designed and simulated through the method of executable description specification and polymorphous parallelism DSE (Design Space Exploration). By reasonable configuration of the edge computing environment and constant optimization of typical application scenarios, such as convolutional neural network and processing of image and graphic, we can meet the challenges of network bandwidth, end-cloud delay and privacy security brought by massive data of the IoE. The philosophy of "Edge-Cloud complements each other, and Edge-AI energizes each other" will become a new generation of IoE behavior principle.

Biases within Digital Repositories: The Getty Research Portal

The influence of computer science on disciplines outside of itself has constituted hybrid communities of practice, such as the digital humanities (DH) and digital art history (DAH), the latter being an offspring of the former.[1] The questions asked within these interdisciplinary fields are substantially different from those in core computer science fields, such as systems theory, language theory, and theory of computation.[2] As a result of the interdisciplinary nature of their inquiries, research questions pertaining to DH and the ethical challenges that accompany them are diametrically opposed to the empirical and often unambiguous measures of validity to which computer scientists are accustomed. This article will explore ethical questions pertaining to collections aggregation systems from the perspective of postcolonial scholarship and seek paths towards addressing the ethical questions currently facing many DH projects. To do so, I will use the Getty Research Portal (hereafter, “the Portal”) as a case study of a digital repository and draw on my personal experience as a software developer working on the project, as well as on current research on biases and ethics within DH.

Declarative Loop Tactics for Domain-specific Optimization

Increasingly complex hardware makes the design of effective compilers difficult. To reduce this problem, we introduce Declarative Loop Tactics , which is a novel framework of composable program transformations based on an internal tree-like program representation of a polyhedral compiler. The framework is based on a declarative C++ API built around easy-to-program matchers and builders, which provide the foundation to develop loop optimization strategies. Using our matchers and builders, we express computational patterns and core building blocks, such as loop tiling, fusion, and data-layout transformations, and compose them into algorithm-specific optimizations. Declarative Loop Tactics (Loop Tactics for short) can be applied to many domains. For two of them, stencils and linear algebra, we show how developers can express sophisticated domain-specific optimizations as a set of composable transformations or calls to optimized libraries. By allowing developers to add highly customized optimizations for a given computational pattern, we expect our approach to reduce the need for DSLs and to extend the range of optimizations that can be performed by a current general-purpose compiler.

Implementation and Hardware-In-The-Loop Simulation of a Magnetic Detumbling and Pointing Control Based on Three-Axis Magnetometer Data

The subject of this work is the implementation and experimental testing of a purely magnetic attitude control strategy, which can provide stabilization after the deployment and pointing of the spacecraft without any attitude information. In particular, the control produces the detumbling of the satellite and leads it to a desired attitude with respect to the direction of the Earth magnetic field, based on the only information provided by a three-axis magnetometer. The system is meant to be used as a backup solution, in case of failure of the primary strategy and is designed considering the constraints set on time of operations, power consumption, and peak electric current for a typical CubeSat mission. The detumbling and pointing algorithms are implemented on the FPGA core of a CubeSat on-board computer and tested by Hardware-in-the-loop simulations. The simulation setup includes a Helmholtz cage, recreating the magnetic environment along the orbit, the on-board computer, a MEMS three-axis magnetometer and Simulink software, on which the attitude dynamics is propagated. Test on the real system can provide useful information to select the parameters of the control, such as the gains, to estimate the limits of the system, the time of operations and prevent failures.

Performance heuristics for GR(1) synthesis and related algorithms

Reactive synthesis for the GR(1) fragment of LTL has been implemented and studied in many works. In this work we present and evaluate a list of heuristics to potentially reduce running times for GR(1) synthesis and related algorithms. The list includes several heuristics for controlled predecessor computation and BDDs, early detection of fixed-points and unrealizability, fixed-point recycling, and several heuristics for unrealizable core computations. We have implemented the heuristics and integrated them in our synthesis environment Spectra Tools, a set of tools for writing specifications and running synthesis and related analyses. We evaluate the presented heuristics on SYNTECH15, a total of 78 specifications of 6 autonomous Lego robots, on SYNTECH17, a total of 149 specifications of 5 autonomous Lego robots, all written by 3rd year undergraduate computer science students in two project classes we have taught, as well as on benchmarks from the literature. The evaluation investigates not only the potential of the suggested heuristics to improve computation times, but also the difference between existing benchmarks and the robot’s specifications in terms of the effectiveness of the heuristics. Our evaluation shows positive results for the application of all the heuristics together, which get more significant for specifications with slower original running times. It also shows differences in effectiveness when applied to different sets of specifications. Furthermore, a comparison between Spectra, with all the presented heuristics, and two existing tools, RATSY and Slugs, over two well-known benchmarks, shows that Spectra outperforms both on most of the specifications; the larger the specification, the faster Spectra becomes relative to the two other tools.

When Sensor-Cloud Meets Mobile Edge Computing.

Sensor-clouds are a combination of wireless sensor networks (WSNs) and cloud computing. The emergence of sensor-clouds has greatly enhanced the computing power and storage capacity of traditional WSNs via exploiting the advantages of cloud computing in resource utilization. However, there are still many problems to be solved in sensor-clouds, such as the limitations of WSNs in terms of communication and energy, the high latency, and the security and privacy issues due to applying a cloud platform as the data processing and control center. In recent years, mobile edge computing has received increasing attention from industry and academia. The core of mobile edge computing is to migrate some or all of the computing tasks of the original cloud computing center to the vicinity of the data source, which gives mobile edge computing great potential in solving the shortcomings of sensor-clouds. In this paper, the latest research status of sensor-clouds is briefly analyzed and the characteristics of the existing sensor-clouds are summarized. After that we discuss the issues of sensor-clouds and propose some applications, especially a trust evaluation mechanism and trustworthy data collection which use mobile edge computing to solve the problems in sensor-clouds. Finally, we discuss research challenges and future research directions in leveraging mobile edge computing for sensor-clouds.

RMT: R-matrix with time-dependence. Solving the semi-relativistic, time-dependent Schrödinger equation for general, multielectron atoms and molecules in intense, ultrashort, arbitrarily polarized laser pulses

RMT is a programme which solves the time-dependent Schrödinger equation for general, multielectron atoms, ions and molecules interacting with laser light. As such it can be used to model ionization (single-photon, multiphoton and strong-field), recollision (high-harmonic generation, strong-field rescattering) and, more generally, absorption or scattering processes with a full account of the multielectron correlation effects in a time-dependent manner. Calculations can be performed for targets interacting with ultrashort, intense laser pulses of long wavelength and arbitrary polarization. Calculations for atoms can optionally include the Breit–Pauli correction terms for the description of relativistic (in particular, spin–orbit) effects. Program summaryProgram Title: (RMT) R-matrix with time-dependenceProgram Files doi:http://dx.doi.org/10.17632/3ptyfg2bmx.1Licensing provisions: GPLv3Programming language: FortranNature of problem: The interaction of laser light with matter can be modelled with the time-dependent Schrödinger equation (TDSE). The solution of the TDSE for general, multielectron atomic and molecular systems is computationally demanding, and has previously been limited either to particular laser wavelengths and intensities, or to simple, few-electron cases. RMT overcomes this limitation by using a general approach to modelling dynamics in atoms and molecules which is applicable to multielectron systems and a wide range of perturbative and non-perturbative phenomena.Solution method: We use the R-matrix paradigm, partitioning the interaction region into an ‘inner’ and an ‘outer’ region. In the inner region (within some small radius of the nucleus/nuclei), full account is taken of all multielectron interactions including electron exchange and correlation. In the outer region, far from the nucleus/nuclei, these are neglected and a single, ionized electron moves in the long-range potential of the residual ionic system and the laser field. The key computational aspect of the RMT approach is the use of a different numerical scheme in each region, facilitating efficient parallelization without sacrificing accuracy. Given an initial wavefunction and the electric field of the driving laser pulse, the wavefunction for all subsequent times and the associated observables are computed using an explicit, Arnoldi propagator method.Additional comments including restrictions and unusual features: The description of the atomic/molecular structure is provided from other, time-independent R-matrix codes [1], [2], [3], and the capabilities (in terms of structure) are, in some sense, inherited therefrom. Thus, the atomic calculations can optionally include Breit–Pauli relativistic corrections to the Hamiltonian, in order to account for the spin–orbit effect. However, no such capability exists for the molecular case. Furthermore, the fixed-nuclei approximation is adopted in the molecular calculations (so nuclear motion is neglected). Similarly, all calculations are restricted to the description of a single electron in the outer region, and consequently the study of double-ionization phenomena is not yet within the capabilities of the method. Finally, the parallel strategy employed necessitates the use of at least two (and usually many more) computer cores. As a result, there is no option for serial calculations and, for most realistic cases, a massively parallel architecture (several hundred cores) will be required.Program repository available at: https://gitlab.com/Uk-amor/RMTReferences[1] C. P. Ballance Parallel R-matrix codes, http://connorb.freeshell.org.[2] R-matrix II codes, http://gitlab.com/uk-amor/rmt/rmatrixii.[3] Z. Mašín et al UKRmol+: a suite for modelling of electronic processes in molecules interacting with electrons, positrons and photons using the R-matrix method, Comput. Phys. Commun., accepted, http://dx.doi.org/10.1016/j.cpc.2019.107092.

A hybrid [formula omitted]-[formula omitted] HOC approach for surface tension driven flows in level set framework

In this study, a higher order compact (HOC) finite difference scheme is proposed for the simulation of multiphase flows with surface tension effects by coupling the pure streamfunction formulation of Navier–Stokes (N–S) equations with the level set method. Unlike most of the existing methods for multiphase flows that solve the incompressible Navier–Stokes equations in the velocity–pressure form or the stream-function vorticity form, we recast the governing equations into the biharmonic form of the N–S equations which carries stream-function ψ as the only variable to be computed. The coupling of this formulation with level set (LS) enables the simulation of two phase flows by the core computation of only the single variable ψ free from discontinuities in the whole physical domain. In the process, we also propose a novel hybrid HOC algorithm for post-processing the pressure on either side of the interface including the jump across it. The robustness of this method is established through a series of benchmark problems such as dispersion of capillary waves, static bubble and oscillating bubble at equilibrium in a fluid at rest, and deforming circular drops in shear flows. We compare our numerical results with available benchmark results and excellent match is obtained in all the cases along with accurate representation of overall trends. Our approach is proved to be very efficient in terms of its easy implementation, relatively low computational time and high order of accuracy.

Improved Estimation of Shear-Wave Velocity by Ordered Weighted Averaging of Rock Physics Models in a Carbonate Reservoir

The aim of this paper is to improve estimation of shear-wave velocity in carbonate rocks. The region being studied is an oil field located in southwest Iran, where Upper Cretaceous carbonates act as the producing reservoir. These fracture bearing carbonates make the study of this reservoir complicated. Moreover, the accuracy and the precision of microstructure and pore type can affect the rock physics models by either directing or distracting these models from the typical models. The geological studies (thin sections, scanning electron microscope and core computer tomography scan) show that the reservoir mainly contains large pores with aspect ratios between 0.1 and 0.9. The existence of fractures is also demonstrated by a formation micro-imager and geological analyses. In this study, the applications of two rock physics models are investigated: (1) self-consistent approximation following the Gassmann equation and (2) Xu–Payne method, which has been carried out using the differential effective medium approach. In addition, for improving model predictions, these two rock physics models are fused by the ordered weighted averaging (OWA) aggregation operator that offers the most appropriate results. Furthermore, there are similarities between the three mentioned models based on the estimated values when compared to the measured dipole shear sonic imager log. Ultimately, the analysis and results demonstrate that the OWA model gives the best compatibility with the original well-log data. In other words, the reason why the OWA model provides the best results in the studied carbonates can be related to its optimization algorithm for defining model parameters.