Intelligent transport systems development. Vadim Shmal

economic and social processes, to improve traffic safety, to improve the environmental situation, to reduce the negative impact of the human factor on the quality of management, to increase the attractiveness of railway transport for passengers and cargo owners. Achieving these goals involves solving a large number of tasks. These, in particular, include:

      ■ improving the efficiency of using the existing railway network by more evenly distributing railway rolling stock in time and space;

      ■ improvement of technological, informational and social components of traffic safety;

      ■ providing managers at all levels with the necessary information to make operational and strategic decisions based on modeling and assessing the impact on the transport system of new and modernized transport facilities;

      ■ formation of a rapid response scheme of transport services, which allows to quickly take measures in case of emergencies, adverse weather conditions, etc.;

      ■ creation of monitoring systems for transport infrastructure and traffic conditions, allowing to assess the state of the transport system in real time and predict its changes.

      3 MODERN SCIENTIFIC AND METHODOLOGICAL APPROACHES TO THE ITS CREATION IN RAILWAY TRANSPORT

      To date, there is no unified understanding of what intelligent transport systems are. In many publications and speeches, they are more or less identified with conventional automated transport systems. An important feature of ITS, which makes it possible to distinguish such systems into a separate class and even into a separate area of research in railway science, is the formal logical and mathematical tools used to solve problems from the standpoint of a system-wide approach to the analysis and management of all systems and processes in railway transport.

      It should be emphasized that modern railway transport belongs to the category of extremely complex technical and organizational systems, the management of which is currently practically impossible within the framework of previously established traditional approaches. The complexity of the transport infrastructure and its facilities (railway junctions, stations, transport corridors, etc.) fundamentally excludes the possibility of working in a fully automatic mode. In other words, it is impossible to effectively manage such a system only with the involvement of classical methods for solving complex mathematical modeling problems, search and development of new approaches are required. At the same time, great hopes are placed on intelligent systems that, along with accurate mathematical models, use data and knowledge accumulated in the course of their activities. The work of such systems can, and sometimes should, be based on the formalized experience of highly qualified specialists. Proceeding from this, JSC «Russian Railways» now needs to develop the fundamental foundations for the creation of intelligent railway systems using complex interdisciplinary approaches that can find practical application in a short time.

      Special attention should be paid to the fact that railway transport management systems, as well as complex systems in general, are characterized by fundamental inaccuracy and uncertainty in both data and management decisions. This makes it possible to attribute such systems from a mathematical point of view to the class of incorrect tasks and makes it possible to evaluate the quality of technical and managerial decisions in a different way. In this case, the promptness of the decisions taken plays a greater role than their optimality, understood in a strict mathematical sense. This quality is an important property of intelligent systems [14,15,16].

      In recent decades, there has been an active development and research of formal methods of working with uncertain data. Until recently, probability theory was the main instrument for accounting for uncertainty. However, the axiomatic limitations associated with it do not allow us to adequately apply probabilistic approaches to solving many important problems in which uncertainty has a different nature or properties. For example, the uncertainty of the events under consideration does not always have a frequency character, objective difficulties often arise with the formalization of a specific probability space, in many cases assumptions about the additive nature of a probability measure are difficult to explain, and sometimes simply unacceptable. For these reasons, at present, along with probability theory with its developed mathematical apparatus, new theoretical approaches to the description of uncertainty and incompleteness of information are actively being investigated. Here, first of all, we should mention the Dempster – Shafer theories, possibilities, interval averages, monotone measures. These theories have less rigid axiomatics, which allows, along with the frequency interpretation of events, to describe events whose uncertainty may be subjective (for example, the probability is determined by a number reflecting the subjective degree of confidence in the event), or in which the number of observed realizations does not allow obtaining reliable conclusions in a statistical sense.

      An important area that can have real practical application in the railway industry when creating ITS is the development of expert systems, i.e. computer programs that can fully or partially replace a specialist expert in some, as a rule, rather narrow problem area. Expert systems began to be developed by artificial intelligence researchers in the 1970s, and already in the 1980s they found their commercial applications. Expert systems function mainly together with knowledge bases, which are a set of facts and rules of logical inference in the chosen subject area of activity. This allows, in general, to model the behavior of experienced specialists in a certain field of knowledge using logical inference and decision-making procedures.

      A person, unlike a computer, has fuzzy thinking, effectively operates with variables not only quantitative, but also qualitative. Therefore, expert systems that model the style of human reasoning are especially successfully used in solving complex problems associated with the use of hard-to-formalize knowledge. It is important to understand that the creation of a specific expert system is a long and expensive process that requires the involvement of specialists in various fields – programmers, knowledge engineers, experts in the field of application under consideration. One of the main problems in this case is the formation of knowledge, which is transmitted during numerous interviews of a knowledge engineer and an expert in the subject area. The stage of knowledge acquisition is one of the main bottlenecks in the technology of creating expert systems due to the low rate of filling the system’s knowledge base. It should be added to this that there are subject areas for which it is often difficult to find an experienced expert person, and sometimes there simply does not exist one. In addition, it has long been noticed that not all experts are ready and able to share their knowledge [2,8.10].

      An important quality of technical systems that allows them to be classified as intelligent is the presence of such properties as:

      ■ learnability – the ability to generate new knowledge and data (models, decision rules) based on inductive inference mechanisms, generalization of statistical data, etc.;

      ■ classification ability – the ability of the system to independently differentiate control objects, environmental influences, control signals, automatically structure data;

      ■ adaptation – the ability of the system to adapt to the changing conditions of the operating environment, correctly take into account the non-stationarity of control data, etc

      One of the promising approaches to the creation of intelligent systems may be to attract the ideas of situational management as a system – wide approach based on formal methods of theoretical artificial intelligence – logical-linguistic models, models of learning technical systems in the construction of management procedures for current situations, deductive systems for building multistep solutions, etc. In this important area of research, as well as in the development of general methodology, theoretical foundations and specific applications, priority undoubtedly belongs to Russian scientists.

      The problem of industrial implementation of intelligent information systems capable of processing data with their inherent a priori uncertainty in railway transport is becoming more and more urgent. In many cases, the data is not only inaccurate and uncertain, but also incomplete, and sometimes unreliable. The