**1. Introduction**

Today, artificial intelligence (AI) has confidently entered our lives. The first mention of it belongs to the mid-50s of the last century. Under AI, we usually understand it as the branch of computer science devoted to develop data processing systems that perform functions normally associated with human intelligence, such as reasoning, learning, and self-improvement (ISO/IEC 2382-1:1993 Information technology–Vocabulary–Part 1). According to this, over the decades, AI has found its application in expert systems supporting decision-making, in heuristic classification, computer vision, pattern recognition, understanding natural language, etc. [1–14]. Here, under AI systems (AIS), we understand systems that include data processing systems that perform functions by AI, in particular by modeling and logic reasoning.

Note. System is a combination of interacting elements organized to achieve one or more stated purposes (according to ISO/IEC/IEEE 15288 "Systems and software engineering–System life cycle processes").

If the modern human brain already possesses skills of adaptation to conditions of various uncertainties in the world around, artificial intelligence systems require creation of effective methods for cognitive solving actual practical problems. "Cognitive solving" means relating to or involving the processes of thinking and reasoning (Cambridge English Dictionary). The applicable mathematical methods are focused mainly on conditions of actions in the logician "if …, that …" according to the gathered information, and on an estimation of traced situations by a manoperator. At increase and expansion of uncertainty conditions, quite often, there are failures and errors because of complexity. It means that search of new methods for advanced solving of AIS practical problems today is very important.

symposiums, conferences, ISO/IEC working groups, and other forums since 2000 in Russia, Australia, Canada, China, Finland, France, Germany, Italy, Kuwait, Luxembourg, Poland, Serbia, the USA, etc. The software tools were awarded by the Golden Medal of the International Innovation and Investment Salon and the International Exhibition "Intellectual Robots," acknowledged on the World's fair of information technologies CeBIT in Germany, noted by diplomas of the Hanover

*Probabilistic Methods for Cognitive Solving of Some Problems in Artificial Intelligence Systems*

Note. The proposed methods below do not replace existing methods for robots actions (for example, the methods of solving the systems of differential equations, the methods of refreshed linear and geometric algebra, geometry, Lie groups, linearization, solving Jacobians and Hessians, Kalman filters, Lyapunov analysis, the methods of biomechanics, graph theory, Laplas transforming for large-scale

The structure of the chapter research is shown in **Figure 1**. It provides an explanation of the essence of cognitive solving of problems on the base of probabilistic modeling, selection of some author's probabilistic models applicable for cognitive solving problems 1 and 2, the practical steps to solve these problems, and five practical examples demonstrating system planning the possibilities of functions performance by using robot-manipulators (in space), by AIS for a coal company and by AIS used for a security service of floating oil and gas platform, example of forming input for probabilistic modeling from monitored data and example of robot route optimization under limitations on risk of "failure" in conditions of uncertainties. Various areas of the examples' applications have been chosen purposely to demonstrate universality and analytical usefulness of the proposed methods and models. Appendices includes the proof for the proposed model of a quite general technology of periodical diagnostics of system integrity and some short models

Industrial Exhibition and the Russian exhibitions of software.

dynamic systems, etc.) [1–14].

*DOI: http://dx.doi.org/10.5772/intechopen.89168*

**Figure 1.**

**5**

*The structure of the research.*

results to estimate quality of used information.

In the present chapter, various AIS for supporting decision-making in intellectual manufacture and robotics systems are analyzed. According to robotics, it is supposed that AIS may be used for solving multiple aerial, land, underground, underwater, universal, and special problems of creation and operation. At the same time, we would like to emphasize that the main efforts of this chapter are not focused on illustrating the capabilities of AIS, but on demonstrating the applicability of author's probabilistic models and methods to improve some of the existing capabilities of AIS.

For this goal, the problem of planning the possibilities of functions performance on the base of monitored information and the problem of robot route optimization under uncertainties limitations are chosen. The choice of these problems in AIS applications is caused on the one hand by increase of quantity and a variety of specific uncertainties conditions, and on the other hand by an urgency and width of areas for their practical use. However, some relevant problems (such as the problems of robotics orientation, localization and mapping, information gathering, the perception and analysis of commands, movement and tactile, realizations of manipulations, and also rational control) for which different probabilistic methods are also applicable have been left out of the scope of work.

For cognitive solving and improvements by the use of probabilistic methods, the chosen problems are transformed more specifically to:


The proposed methods for cognitive solving AIS problems are based on theoretical and practical author's researches [15–37] and need to be used either in combination or in addition to existing methods. There, where often it is required prognostic analysis or where the used approaches are not effective, the proposed methods can be used as rational basis or alternative.

The proposed and referred author's methods and models can be used in AIS life cycle to form system requirements, compare different processes, rationale technical decisions, and estimate reliability, quality, and risks. The decisions, scientifically proved by the offered models and software tools, can provide purposeful essential improvement of quality and mitigation of risks and decrease expenses for created and operating systems. The spectrum of the explored systems by these methods includes systems (not only AIS) operated by government agencies, manufacturing structures (including power generation, coal enterprises, oil and gas systems), food storage, space industry, emergency services, municipal economy, etc. [15–19, 22–37]. The supporting software tools are original Russian creations registered by Rospatent [38–44]. They have been presented at seminars,
