Software & Systems

Articles of journal № 3 at 2017 year.

Order result by: Public date | Title | Authors

21. A spatio-temporal model of a material and technical resources traffic network of a radio-electronic facilities group exploitation [№3 за 2017 год]
Authors: S.V. Ignatev, V.B. Tikhonov, A.V. Krasnikov , А.А. Оsipov
Visitors: 7694
Support of complicated radio-electronic facilities (REF) reliability at the established level mostly depends on the operation effectiveness of the current technical exploitation system, with the logistic system (LS) as one of the component part. The main task of the LS is considered to be well-timed supply of material resources (MR), which are necessary for REF proper exploitation. At the same time, the task of fast and proper decision-making when controlling necessary resources supply becomes relevant due to considerable space diversity and remoteness of special-purpose systems with REF as the basic element from storage and maintenance bases. To perform the task it is necessary to know possible (preferable) routs of MR delivery, taking into consideration emergency situations, traffic jams, roadways heavy maintenance, natural disasters and other factors, as well as the choice of optimal route variants according to the desired time value and/or MR delivery cost. Taking into account the above-mentioned facts, this research suggests a spatio-temporal model of logistics system of the group of radio-electronic facilities transportation network exploitation. This model can help to solve the task of selecting preferable MR delivery routes according to established criteria.

22. Anaphor resolution system development based on machine learning methods [№3 за 2017 год]
Authors: A.V. Sokolov, Batura T.V.
Visitors: 8577
The paper proposes and implements a method for the anaphora resolution of third person pronouns in Russian texts. The problem of finding the true pairs “anaphor-antecedent” is considered as a binary classification problem. Initially, the authors perform morphological and syntactic analysis of the text. The morphological analyzer used the pymorphy2 library. The parsing has been performed using MaltParser. The algorithm of anaphora resolution itself consists of three stages. First stage includes searching for all pronouns, then there is a compilation of many potential antecedents, and finally the most suitable candidate is selected. The component of creating a set of candidates for antecedents is based on using distance, morphological and syntactic filters. Classification uses the Random Forest algorithm. The anaphoric classifier takes into account 78 different features. The authors performed a series of experiments in order to prove the effectiveness of the proposed method. They showed that that the quality of the analyzer improves if we do not take into account the morphological case. It can also be noted that the number of trees taken for calculation has a lesser effect on the final result when taking a feature set without cases. The paper considers the main difficulties in developing the anaphora resolution systems. First, the search for anaphoric relations is in the semantic domain, and therefore it is difficult to formalize. Second, there are some features of the Russian language, such as developed morphology, morphological and syntactic ambiguities, which adversely affect the result.

23. Development of a pattern recognition system for a mobile robot [№3 за 2017 год]
Authors: Ulyanov, S.V., A.G. Reshetnikov, K.V. Koshelev
Visitors: 9595
The main development source of robotic systems and artificial intelligence systems is improvement of computer technology, devices and sensors that allow obtaining information about the environment. This process is inextricably linked with the development of software on-board processors, which are used in autonomous robot managing. The article gives a brief description of a module of a developed image recognition system for an autonomous robot, which is based on a stereo vision technology. Nowadays, more and more articles mention the systems consisting of combinations of several video cameras, laser radars, ther-mal imagers and other sources of information about the environment. Equipment of this class became widespread when creating au-topilots for vehicles, security control systems, unmanned aerial reconnaissance, various household robots. Stereo vision allows ob-taining data on image depth, the distance to objects. It also provides an opportunity to build a three-dimensional picture of the sur-rounding world. The paper describes the configuration of the equipment used (a pair of cameras fixed on a common base): a stereopair calibrating process, the calibration pattern on the image was searched using the ROS framework. Images obtained from cameras are subjected to rectification (a transformation when both images are projected onto a plane parallel to the line that connects the optical centers of the cameras). The article considers the fundamentals of projective transformations and epipolar geometry. The paper presents the idea that un-derlies the construction of depth and displacement maps. As a result, it presents a recognition software module developed using the Python programming language.

24. Calculation of the optimal dependency of information resource price during its marketing at several markets [№3 за 2017 год]
Authors: S.A. Amelkin , O.S. Ivanova
Visitors: 3788
The paper considers the problem of determining the pricing policy of the company that implements a market information resource. One of the key challenges of the information market – information cost determination. However, this raises a number of difficulties due to the nature of the goods. The paper considers the information goods as a certain software product. The sale of information resources is characterized by the fact that replication of this resource can be without any significant additional costs. This leads to increasing in resource volume during its dissemination. At the same time, price setting is one of the key factors determining further distribution and promotion of goods, which takes into account a variety of factors including the way it is distributed. The paper solves the problem of determining the optimal dependence of the information resource price on time for the case when the retailer has no competitors on the market, the market is saturated, and buyers do not have the possibility of further dissemination of the resource. The authors obtain the conditions for optimal selection of the information resource price, calculate the maximum possible income of the company for a linear demand function, and describe the conditions governing the duration and volume of sales ensuring a high average company income. When the function of demand is known (it can be found using an approximation of the experimental data), the obtained conditions allow selecting the optimal change in sell prices as a function of the number of sold copies of the information resource.

25. Channels implementation of ARINC 653 specification in RTOS Baget 3 [№3 за 2017 год]
Authors: A.N. Godunov , V.A. Soldatov, I.I. Homenkov
Visitors: 9252
The article considers ARINC specification 653 channels, which are intended for interprocess communication. The paper proposes the basic methods of designing channel driver, as well as their interaction with the Russian real-time operating system Baget 3 (RTOS Baget 3). RTOS Baget 3 user interface is based on ARINC 653 Specification and POSIX standard. The channel ac-cess is performed through ports. Every channel has the only port for sending messages and a single or several destination ports for receiving messages. Messages can be sent through a channel in only one direction. An application program can use the channels in the processes that correspond both ARINC and POSIX. The paper specifies the requirements for channel drivers and describe notification control means used for driver development. It also considers the stages of initialization, sending/receiving messages and implemented algorithms. The proposed methods of im-plementing drivers interaction with RTOS allow avoiding data recording to the other process memory that significantly enhances the system reliability. Distributed computing systems use the same interface when separate subsystems are interconnected by means of shared access buses (VME, RapidIO, Fibre Channel) or Ethernet. The connection table for processor units is specified at the stage of system configuration. The channel driver, designed for RapidIO communication environment that offers fast operation speed, is used in a multiprocessor digital signal processing system.

26. Optimization problems solution based on superelement modeling of oil-field development [№3 за 2017 год]
Authors: I.V. Afanaskin , P.V. Yalov, Giatsintov A.M., A.V. Roditelev
Visitors: 6850
Oil fields of Russia are mainly developed by waterflooding. Most of them are in the 3rd or 4th stage of development. Consequently, the water cutting of well production is 80–90% or more. In these conditions, in order to optimize the development of deposits, oil engineers try to reduce water production and injection while maintaining or increasing oil production. For this purpose, there are the tasks of field development control and regulation. These problems are solved using various mathematical models. This paper considers a superelement mathematical waterflooding model based on a two-phase filtration model for weakly compressible immiscible liquids (oil and water) in an elastic bed under the Darcy law. The system of differential equations for pressure and saturation is approximated on Voronoi diagram in an entirely explicit manner. The size of the superelements is comparable to the distance between the wells. This allows performing calculations without using special software. To solve inverse problems (determine model coefficients or optimize development parameters), the work uses Newton's method and the conjugate gradient method. In the classical setting of an inverse problem, the optimization theory methods should be applied directly to the mathematical model of the process under study. However, when solving oilfield development problems, the number of optimization parameters can be large, and the complexity of the mathematical model is quite high. Therefore, the application of optimization theory methods directly to a mathematical model can be very time-consuming. To overcome this contradiction, it is proposed to build statistical dependencies of the development indices on the required parameters using a mathematical model of waterflooding, and then to apply optimization theory methods no longer to the mathematical model, but to the statistical dependencies obtained. To illustrate this approach, we consider the solution of the problem of model adaptation to absolute permeability. It is established that the application of the conjugate gradient method directly to the waterflooding model gives an error in determining the permeability of 11,8 %. Applying the same method to a statistical dependence of a model adaptation error (on accumulated production and injection of oil and water) on the logarithm of permeability gives an error in determining permeability of only a little more, it is 15 %.

27. Modern trends in automated usability evaluation and behavioral factors in search engine algorithms [№3 за 2017 год]
Author: M.A. Bakaev
Visitors: 6649
Nowadays, there are persisting ambiguities in usability conceptualization for quantitative measurement and a certain gap between its more objective components (effectiveness and efficiency) and a fairly subjective aspect of user satisfaction. The paper mentions that the practical goal is not necessarily measuring website’s usability, but rather improving its quality-in-use or assessment by respective algorithms of major search engines, which increase the importance of “behavioral factors” in their rankings. The paper describes traditional methods in usability evaluation and design, as well as up-to-date tools that introduce certain automation into them. The authors identify approaches within automated usability evaluation and summarize possibilities and limitations of each one. The approaches are based on interactions, metrics and modeling. Furthermore, there is an overview of hybrid approaches that use more and more AI and machine learning methods. Therefore, an effective method for website interaction quality evaluation is neural networks, which are applied within evolutionary web design algorithms to calculate fitness function for candidate solutions. The final part of the paper describes how global search engines (such as Yandex and Google) perform automated website usability evaluation and provides a list of the most important behavioral factors. In fact, they mostly consider usability as a “black box” and actively apply machine-learning methods to approximate its calculation. Thus, usability conceptualization and practical assessment by search engines, which nowadays shape the web to a great extent, is constantly evolving and, in our opinion, calls for respectively flexible approaches from web designers and engineers.

28. Comparison of russian and foreign microprocessor performance [№3 за 2017 год]
Authors: N.D. Baykov , A.N. Godunov
Visitors: 10727
This article provides performance comparison for three different MIPS processors (RM7000, XLP316 and 1890VM8YA) and describes their architecture details. To compare processor performance, the testing technique is developed and implemented using C and Assembler. The technique consists of three consecutive stages. At the first stage the authors measure processor instructions execution time providing that both instructions and data required are already stored in a primary cache. The instructions are divided into several groups. There are the results for members of each group. Primary and secondary cache efficiency is benchmarked on the second stage. The article provides the results for secondary cache and RAM access time only. The third stage uses synthetic performance tests. The obtained results are compared with theoretical estimations based on the results of the first two stages. The advantage of the proposed technique is in its independence from compiler and operating system specified. All measurements are carried out in clock cycles using special purpose coprocessor registers.

29. A framework for analysis and forecasting of time series in the development of proactive decision support system components [№3 за 2017 год]
Authors: M.V. Shcherbakov, K.S. Zadiran, A.V. Golubev, Al-Gunaid Mohammed Amin
Visitors: 7853
The article describes the developed framework for automation research on software design for decision support proactive systems. In particular, it considers the problem of analysis and forecasting time series to form automation components in predicting various processes. In practice, there are various data analysis libraries (R, Python) used for prototyping the component uses. The main problem is a lack (or insufficiency) of implementation of analysis methods, i.e. the sequence of actions during analysis. In addition, there is a number of limiting factors that affect the efficiency of analysis and forecasting components design: (i) time-consuming routine operations during time series manual analysis; (ii) developer’s insufficient qualifications as a deterrent in implementing proactive system components; (iii) a frequent need in the analysis of a large number of similar data. The proposed software solution (a software in Python) automates the process of time series analysis and generate a LaTeX report. Formation is automatic in accordance with CRISP-DM methodology. The report provides a comprehensive analysis of time series in accordance with published and accepted methods. All functions of the framework can be attributed to one of the following groups: data loading and the internal data frame; descriptive time series analysis with visualization; prediction and report making. The paper shows an example of using a framework to solve the problem of proactive management decisions making support in energy management systems (forecasting of electricity consumption).

◄ ← Preview | 1 | 2 | 3