The Department of Intellectual and Information Systems of the Taras Shevchenko National University of Kyiv conducts scientific research on theoretical and applied aspects of artificial intelligence. The department is designing principles, models, methods and tools contributing to the development of the theory of artificial intelligence and its practical applications.

The studies are based on connectionist and evolutionary approaches.  The theoretical direction includes new methods for solving clustering tasks, restoring missing values in data tables, predicting, continuous and discrete optimization. The advantage of using hybrid technologies of artificial intelligence with an application of neural networks, evolutionary modeling and elements of the fussy set theory is proved. At the same time the process of solving these tasks is considered to be an evolutionary process with elements of random but directed search. The peculiarity of the results is their application with incomplete, unreliable and ambiguous data, as well as data obtained in the form of expert conclusions. The development of technologies for processing natural language information is being carried out, namely, the creation of models and methods required for constructing systems of tests’ automatic generation, potential responses to them, and appropriate evaluation. Such systems are obligatory for distance and self-education and may be a structural part of “lifelong education” concept. Simultaneously the work is underway to create automated systems of education and module knowledge control, the work with which is carried out according to study individual trajectory.

A peculiarity of these systems’ application is an optimized and individualized selection of study materials and evaluation schemes. Methods for assessing the adequacy of knowledge control with possible adjustment of the questions complexity are also offered.

A significant part of the research is carried out in the applied direction of “ensuring a safe human environment”. Since a lot of people die as a result of car accidents, fire and chemical disasters, a number of tasks have been solved minimizing both human casualties as well as material damage. These tasks include: optimization of the fire truck route to the place of fire; defining the fire safety level of the building, establishment of reasonable housing price, based on its fire safety level; optimization of fire monitoring systems in large buildings and constructions; determination of possible ways and time of the fire spread to an especially dangerous object,  calculation of the hazardous chemical concentration in the post-accident period and its post-forecasting.

Most of these tasks are solved under conditions of uncertainty and critical time for decision-making. In offline mode appropriate models are constructed, based on

neuro-fuzzy constructs. Their parametric and structural optimization is carried out by means of developed evolutionary methods, since the offered models are poly-extreme, non-smooth dependencies. In the event of a fire or an accident (online mode), the original data is introduced into the model, calculations are carried out, the obtained results are the basis for the decision-making by the persons in charge. In some tasks production and logical models of knowledge are additionally used, allowing correcting and specifying the results of forecasting and clustering.

It is planned to develop robotic systems with elements of computer vision capable of learning and self-learning on the basis of the connectionist approach and the use of traditional logics.