Approach to knowledge management and the development of a multi-agent knowledge representation and processing system

Objectives. A multi-agent knowledge representation and processing system (MKRPS) comprises a distributed artificial intelligence system designed to solve problems that are difficult or impossible to solve using monolithic systems. Solving complex problems in an MKRPS is accomplished by communities of intelligent software agents that use cognitive data structures, logical inference, and machine learning. Intelligent software agents are able to act rationally under conditions of incompleteness and ambiguity of incoming information. The aim of the present work is to identify models and methods, as well as software modules and tools, for use in developing a highly efficient MKRPS.

Methods. Agent-based modeling methods were used to formally describe and programmatically simulate the rational behavior of intelligent agents, expert evaluation methods, the mathematical apparatus of automata theory, Markov chains, fuzzy logic, neural networks, and reinforcement learning.

Results. An MKRPS structure diagram, a multi-agent solver, and microservices access control diagram were developed. Methods for distribution of intelligent software agents on the MKRPS nodes are proposed along with algorithms for optimizing the logical structure of the distributed knowledge base (DKB) to improve the performance of the MKRPS in terms of volume, cost and time criteria.

Conclusions. The proposed approach to the development and use of intelligent software agents combines knowledge-based reasoning mechanisms with neural network models. The developed MKRPS structure and DKB control diagram includes described methods for optimizing the DKB, determining the availability of microservices used by the agents, ensuring the reliability assurance and coordinated functioning of the computing nodes of the system, as well as instrumental software tools to simplify the design and implementation of the MKRPS. The results demonstrate the effectiveness of the presented approach to knowledge management and the development of a high-performance problem-oriented MKRPS.

1. Zaytsev E.I., Khalabiya R.F., Stepanova I.V., Bunina L.V. Multi-Agent System of Knowledge Representation and Processing. In: Kovalev S., Tarassov V., Snasel V., Sukhanov A. (Eds.). Proceedings of the Fourth International Scientific Conference “Intelligent Information Technologies for Industry” (IITI’19). Advances in Intelligent Systems and Computing. Springer; 2020. V. I. P. 131–141. https://doi.org/10.1007/978-3-030-50097-9_14

2. Baranauskas R., Janaviciute A., Jasinevicius R., Jukavicius V. On Multi-Agent Systems Intellectics. Inf. Technol. Control. 2015;44(1):112–121. https://doi.org/10.5755/j01.itc.44.1.8768

3. Darweesh S., Shehata H. Performance Evaluation of a Multi-Agent System using Fuzzy Model. 2018 First International Workshop on Deep and Representation Learning (IWDRL). 2018. P. 7–12. https://doi.org/10.1109/IWDRL.2018.8358208

4. Russel S., Norvig P. Iskusstvennyi intellekt: sovremennyi podkhod. T. 2. Znaniya i rassuzhdeniya v usloviyakh neopredelennosti (Artificial Intelligence: A Modern Approach. V. 2. Knowledge and Reasoning under Uncertainty): transl. from Engl. St. Petersburg: Dialektika; 2021. 480 p. (in Russ.).

5. Russel S., Norvig P. Iskusstvennyi intellekt: sovremennyi podkhod. T. 3. Obuchenie, vospriyatie i deistvie (Artificial Intelligence: A Modern Approach. V. 3. Learning, Perception, and Action): transl. from Engl. St. Petersburg: Dialektika; 2022. 640 p. (in Russ.).

6. Sutton R.S., Barto E.J. Obuchenie s podkrepleniem (Reinforcement Learning): transl. from Engl. Moscow: DMK Press; 2020. 552 p. (in Russ.).

7. Graesser L., Keng W.L. Glubokoe obuchenie s podkrepleniem: teoriya i praktika na yazyke Python (Reinforcement Learning: Theory and Practice in Python): transl. from Engl. St. Petersburg: Piter; 2022. 416 р. (in Russ.). [Graesser L., Keng W.L. Foundations of Deep Reinforcement Learning: Theory and Practice in Python. Addison-Wesley Professional; 2020. 416 p.]

8. Pumperla M., Ferguson K. Glubokoe obuchenie i igra v go (Deep Learning and the Game of Go): transl. from Engl. Moscow: DMK Press; 2020. 372 p. (in Russ.).

9. Red’ko V.G. Evolyutsiya, neironnye seti, intellekt: Modeli i kontseptsii evolyutsionnoi kibernetiki (Evolution, Neural Networks, Intelligence: Models and Concepts of Evolutionary Cybernetics). Moscow: LIBROKOM; 2013. 224 p. (in Russ.).

10. Winder P. Reinforcement Learning. Industrial Applications of Intelligent Agents. O’Reilly Media, Inc.; 2021. 382 р.

11. Houhamdi Z., Athamena B., Abuzaineddin R., Muhairat M. A Multi-Agent System for Course Timetable Generation. TEM Journal. 2019;8(1):211–221. https://doi.org/10.18421/TEM81-30

12. Aly S., Badoor H. Performance Evaluation of a Multi-Agent System using Fuzzy Model. 2018 First International Workshop on Deep and Representation Learning (IWDRL). 2018. P. 175–189. https://doi.org/10.1109/IWDRL.2018.8358208

13. Zaytsev E.I. Method of date representation and processing in the distributed intelligence information systems. Avtomatizatsiya i sovremennye tekhnologii = Automation. Modern Technologies. 2008;1:29–34 (in Russ.).

14. Batouma N., Sourrouille J.-L. Dynamic adaption of resource aware distributed applications. Int. J. Grid Distrib. Comput. 2011;4(2):25–42. Available from URL: http://article.nadiapub.com/IJGDC/vol4_no2/3.pdf

15. Nurmatova E.V., Gusev V.V., Kotliar V.V. Analysis of the features of the optimal logical structure of distributed databases. In: GRID Workshop Proceedings - GRID 2018 Selected Papers of the 8th International Conference “Distributed Computing and Grid-technologies in Science and Education”. Dubna; 2018. V. 2267. P. 579–584. Available from URL: https://ceur-ws.org/Vol-2267/579- 584-paper-111.pdf