Development of an Intelligent System for Managing Virtual Concierges

The paper describes the process of developing an intelligent system for managing virtual concierges (curators). The system allows you to manage virtual curators of health, location, security, and access. Each of the curators is based on a neural network trained with a reinforcement algorithm. When training neural networks, the OpenAI Gym environment and algorithms from the Stable Baselines 3 library were used. To find the shortest path in the building, Dijkstra's algorithm was modified. The work contains: 65 pages of the main text, 3 chapters, 8 tables, 28 sources used, 46 figures, 1 appendix (Appendix A. Listing of the appendix). Chapter 1 provides an analysis of business processes and technologies used, including analysis of existing platforms for implementing multi-agent systems, analysis of machine learning methods, selection of a library for developing machine learning algorithms, and analysis of shortest path algorithms. Chapter 2 contains the design of a system for managing virtual curators. Chapter 3 contains the implementation of the system and the analysis of the results.