Software & Systems

Articles of journal № 1 at 2019 year.

Order result by: Public date | Title | Authors |

11. A constructive learning method for artificial neural networks with weighted rates [№1 за 2019 год]
Author: M.A. Kazakov
Visitors: 5805
The paper proposes a constructive method for teaching artificial neural networks with different correction parameters for neu-rons added at different stages of training. This method allows dealing with local minimization and at the same time controlling a neural network scale. It is assumed that dealing with local minimization might become more effective due to the difference in the learning rate, when the neurons added at later training stages are adjusted more intensively than the neurons added at early stages. The paper presents statistical data obtained in MNIST examples using the proposed method, the standard gradient descent method and the constructive teaching method. There is a Python program that has been developed using the numpy and matplotlib libraries in order to conduct numerical experiments that allow comparing the considered methods. A neural network is a direct distribution network, where the neuron inputs are connected to all the outputs of the previous layer. The activation function for all neurons is an exponential sigmoid. The training has been carried out by the method of back propagation of errors. The sum of squares of distances between output signals and reference values has been used as an evaluation function. The paper describes study conditions in detail and provides a graph illustrating the dynamics of the decline in the evaluation function value for all three methods. It is also assumed that the proposed method will reduce the impact of the learning process on the new data class on the effectiveness of the neural network in the classes that the network has learned at the early stages.

12. Formation of a safety barrier for a spacecraft under spaсe debris impact using fuzzy logic methods [№1 за 2019 год]
Authors: V.K. Kemaykin, I.V. Kozhukhin
Visitors: 4738
The paper proposes the algorithm to form a safety barrier from space debris impact for a spacecraft independent operation. The efficiency of the safety barrier depends on responsiveness, economy and efficiency (reliability) of protective measures that form the safety barrier. Nowa-days, the safety barrier includes measures implemented in a spacecraft with the human operator in an automatic mode. For this purpose, space debris must be detected in time and the time to a predicted collision must be sufficient to make a decision (about 28 hours). Then an operator can evaluate and consider the importance of parameters and form a proper barrier. An automatic avoidance maneuver is performed for other cases when there might be a collision with space debris. This maneuver associates with changes in the orbital parameters of a spacecraft. The task of safety barrier efficiency estimation appears under the conditions of autonomous orbital functioning of a spacecraft. This task should take into account environmental conditions, when safety barrier is formed automatically in real time mode on board. The developed knowledge base of parameter importance allows estimating the safety barrier effectiveness taking into account the situa-tion for each potentially dangerous (by the collision criterion) object in the automatic operation cycle of an on board spacecraft computer. The requirement for spacecraft functioning autonomy and the countering of a possible danger from a space debris impact is based on the use of the fuzzy set theory including the principle of fuzzy merging of objectives and restrictions. Initial estimates of a safety barrier effective-ness implement the principle of a guaranteed result with estimable criteria balance. Real-time estimation of the effectiveness takes into ac-count the importance of safety barrier parameters, which are sets of rules from the knowledge base about the required and existing effective-ness (reliability) restrictions, time and fuel consumption for implementing potential safety barriers. The obtained results show that the effectiveness of potential safety barriers may vary in real conditions depending on the importance of the parameters.

13. The unified representation of LTL and CTL logics formulas by recursive equation systems [№1 за 2019 год]
Authors: Korablin Yu.P., Shipov A.A.
Visitors: 6130
Nowadays, to solve the formal verification problem using the Model Checking method, the following logics are often used: the linear-time temporal logic (LTL), the computation tree logic (CTL) and CTL* that combines the capabilities of both other logics. However, each of these logics has its own disadvantages, limitations and expressiveness problems due to their syntactic and se-mantic features. Therefore, there is no universal temporal logic at the moment. The authors are convinced that special representations, which are based on systems of recursive equations in regard to tem-poral logics, can extend their expressiveness, as well as unify their syntax. Thus, they allow building their common and uniform notation. The paper proposes and considers a special RTL notation that is based on systems of recursive equations and the accus-tomed LTL and CTL semantic definitions. The notation is intended to solve the problem of unification of expressiveness of both logics, which in turn expands the expressiveness each one of them. The unification of their syntactic structures will give an opportunity to develop a uniform approach for the Model Checking problem. The authors provide a detailed definition of the RTL notation; give corresponding axioms and theorems. The paper also rep-resents a number of examples and statements that clearly demonstrate the RTL expressiveness capabilities. The purpose of the paper is to demonstrate key features and capabilities of the RTL notation, which are the basis for the au-thors' further research on solving the problem of system models verification.

14. Development of the concept of data migration between relational and non-relational database systems [№1 за 2019 год]
Authors: Yu.A. Koroleva, V.O. Maslova, V.K. Kozlov
Visitors: 4928
The article investigates relational and non-relational approaches to constructing, storing, and extracting data. Nowadays all information and analytical systems use databases. These systems require the ability to process, read, write specific data sets that need to be organized, structured and stored. Finding a suitable database and database management system is one of the most common problem for many companies, as this choice will determine performance, reliability, security, design features and other work features. Usually several data mod-els can be used in one information system of a company. For example, companies use a relational database for the tasks that re-quire using full data consistency and transaction control, whereas analytical, aggregated or meta-data can be kept in a NoSQL database. This separation is often necessary for the most effective functioning of the final product. Combining these systems is the main problem. The research discovered the most popular database management systems for both approaches of developing databases. Their advantages and disadvantages are analyzed. As the first phase of data preparation for transparent migration between two systems, the authors propose the transformation scheme from relational data to non-relational data. This scheme is based on the databases internal organization and their peculiar properties.

15. Algorithmic and software implementation of a cognitive agent based on G. Polya’s methodology [№1 за 2019 год]
Authors: S.S. Kurbatov, Fominykh I.B., A.B. Vorobev
Visitors: 9013
The paper describes an original approach to creating an integrated problem solving system (cognitive agent). The system in-volves a tight integration of linguistic processing stages, an ontological representation, a heuristically oriented solution and visu-alization. The system concept is based on the Polya’s methodology interpreted in algorithmic and software implementation. The system is implemented in a mock-up version and tested in the subject area “school geometry”. The linguistic component uses the problem canonical description obtaining method through paraphrasing and mapping it into a semantic structure. An automated solution search is based on implementing the rules that reflect the axioms of the respective subject areas. The heuristics presented in the ontology define the rules. The heuristics are designed as semantic network structures, which allows organizing a rule multiple-aspect search and selection justification as a natural language comment. Conceptual (cognitive) visualization provides the solution visual representation by interpreting a text file with information to display graphical objects, as well as comments on the solution process. Comments include natural language descriptions of rules (axioms, theorems), heuristic and empirical justifications for their choice and links to visualized objects. The paper defines experiments that demonstrate visualization possibilities of task drawings and ontology fragments, natural language phrases, mathematical and formal logic formulas. The ontology is implemented in the DBMS Progress. Visualization programs are implemented in JavaScript using JSXGraph and MathJax. The implementation provides a step-by-step solution view in different directions with dynamic changing in drawing and related comments. The authors have interpreted experimental results and planned the study to develop the described approach.

16. Developing a software and hardware system for GSM remote diagnostics of surface vehicles [№1 за 2019 год]
Authors: D.S. Lavygin, V.V. Levshchanov , A.N. Fomin
Visitors: 5406
The paper describes the structure of a software and hardware system for remote diagnostics of surface vehicles by a GSM channel. The developed device is relevant due to the need of operational control and statistical processing of a large amount of diagnostic information. A unified standard, which appeared in the early 2000s, made it possible to diagnose a number of vehi-cles by universal tools without considering design features of each of them. The paper shows the basic scheme of interaction between hardware and software elements of the system and the structural diagram of a diagnostic adapter hardware part. The diagnostic adapter device consists of several modules controlled by two STM32 microcontrollers (STMicroelectronics). One of them controls automobile interfaces and provides data exchange with electronic equipment of a vehicle. The second controller processes and transfers data to a remote server. At the same time, the obtained diagnostic data is linked to a vehicle geographic location (calculated using the built-in GPS module), which allows identifying the relation between operating conditions and a number of faults that occur. The paper also shows a detailed scheme of software interaction between hardware and software components of the com-plex, and describes interaction principles between the vehicle diagnostic adapter and a remote server for collecting, storing and processing diagnostic information. The software implementation of the server component includes two modules united by a common database. The first module provides a user-friendly web interface. The second one implements diagnostic algorithms.

17. The development of a remote client for automated data transfer in UNIX-based systems [№1 за 2019 год]
Authors: E.V. Palchevsky, A.R. Khalikov
Visitors: 4776
This paper is devoted to the development of a hardware-software module for UNIX-based systems, called RCM (Remote Client Management). The module provides data transfer within the Protection hardware-software complex protecting from DDoS-attacks. The main RCM features are high-speed data processing and protection from DDoS attacks based on neural networks. The paper discusses the problem of processing software data and substantiates the need for a mathematical analysis to iden-tify new self-learning methods of neural networks. The paper also presents the developed self-learning neural networks neces-sary for data transmission and protection from DDoS attacks. The developed method for self-learning a neural network is based on combining signal and differential learning methods. Therefore, the neural network can quickly learn in a short time. The functionality of the developed remote client allows managing this module both through the web interface and the console mode. Testing of the developed software in the combat mode has shown the load values for computer resources. Long-term testing of RCM has shown quite a low load on the central processor and solid-state drive during DDoS-attacks. Naturally, optimal load allows not only processing large information flows, but also provides the possibility of parallel launch of resource-intensive computing processes without any disruption to the operating system operation. The testing has been carried out on the computational cluster servers (together with APK “Protection”) in one of the Moscow data centers, where RCM performed stably.

18. Computer simulation of physical interactions of technical surfaces at the micro-level [№1 за 2019 год]
Authors: A.A. Rachishkin, Bolotov A.N., O.V. Sutyagin
Visitors: 4527
The paper presents the software design for computer simulation of physical interactions of engineering surfaces at the micro level. It describes general principles of the system and analyses several simulated physical processes. It also considers contact interaction of rough surfaces including those with functional coatings. It calculates frictional parameters under various operating conditions and the thermal resistance of microindented joints of various machine parts. A flexible algorithm of the program and an independent segmented mathematical logic are developed for the optimal calcu-lation of simulated processes using numerical methods. Input data is based on micro-geometric and physical-mechanical proper-ties of real surfaces. This allows adjusting the model considering most technical surface properties. Developing software for simulation of contact interaction of technical surfaces makes it easier to solve engineering tasks and reduces the amount of re-sources needed for research. The software is divided into separate modules. The definition of general principles for each module development is benefi-cial for the horizontal scaling of a system for modeling physical interactions. This approach allows optimizing the number of in-put and output parameters, adding and modifying existing algorithms and configuring only necessary modeling processes. A modular structure optimizes the software development process. Due to the ability to remove and modify individual program modules without affecting the overall structure, it is possible to solve a wide range of engineering tasks effectively. The paper gives some examples of module operation for contact and friction interactions, as well as modeling thermal re-sistance of joints. The general algorithmic logic and the inherited surface topography data allow modeling these physical processes as real as possible.

19. Flexibility of using input and output parameters of standard and non-standard functions in MatLab [№1 за 2019 год]
Author: O.G. Revinskaya
Visitors: 8066
Based on a review of recent papers, the paper reveals the contradiction between the understanding of the breadth and flexibility of using input and output parameters of standard functions and the feeling of rigid predetermination when describing and using similar parameters of non-standard MatLab functions. This contradiction is resolved by a detailed analysis of the capabilities provided by MatLab (including its latest versions), so that the function parameters (when it is called) are interpreted as mandatory or optional, positioned or unpositioned, typed or untyped, etc. This variety of properties of input and output parameters provides flexibility in the application of standard MatLab functions. It is shown that by default MatLab controls only formal excess of the number of parameters used when calling a function (standard, non-standard) over the number of corresponding parameters specified in its description. For the parameters of a non-standard function to have certain properties, it is necessary to organize a function body program code in a special way: to check how many parameters are specified when the function is actually called, what type of information enters the function and exits through parameters; to analyze which optional parameters are set and which are not, etc. Such organization of the function body has been remaining very laborious for a long time. Therefore, the latest versions of MatLab have standard functions that auto-mate some of the performed operations. Thus, the article systematizes a set of measures that allow the parameters of a non-standard function to have the same breadth and flexibility of use as the parameters of standard MatLab functions. Based on personal experience in applied programming and teaching MatLab, the author shows simple examples that illus-trate in detail how to write non-standard functions with parameters that have the appropriate properties.

20. Simulation of messages transmission between moving objects in a transport environment [№1 за 2019 год]
Authors: S.V. Rudometov, O.D. Sokolova
Visitors: 4489
In recent years, there have been a lot of research in the development of wireless networks connecting vehicles. To connect ve-hicles to each other, as well as to connect them with roadside equipment, there are wireless networks that have are based on the principle of mobile ad hoc networks (Vehicular Ad Hoc Network (VANET)). The paper considers modeling of traffic in the transport network segment and data transmission from one node located on a moving object to other traffic participants. To simulate the movement of vehicles and message transmission, the Manufacturing and Transportation Simulation System (MTSS) is used. The system has been developed earlier by one of the authors. The MTSS allows visual building of simulation models of technological systems and conducting various simulation tests with these models. The paper considers MTSS simulation. The MTSS includes transceiver devices installed in moving objects (cars) or in sta-tionary objects located along the route. The authors study data transfer in this network - message transmission between cars (for example, an emergency message). The paper considers two options of a transport network segment that are: straight-line (high-way) and a square with road intersections. Experimental data show that the interference role in message transmission in a straight-line segment is not as significant as in a segment with road intersections.

◄ ← Preview | 1 | 2 | 3 | Next → ►