THE USE OF AN ADAPTIVE RL AGENT FOR DYNAMIC REGULATION OF ECONOMIC FLOWS IN AN MMO SIMULATION

Figurov Maxim Olegovich, Mishchuk Bogdan Rostislavovich

Regulation of economic flows in closed virtual economies is one of the most urgent tasks in the modern field of information technology. Unlike in the real world, the value of a virtual currency directly depends on the proper configuration of its sources of origin and distribution, called faucets and sinks. Mistakes in regulating economic flows lead to economic destabilization, a sharp spike in inflation or deflation, and the collapse of the virtual market, which in the long run leads to a decrease in the number of users and large financial losses. Traditionally, the balance of the virtual economy is handled by a whole department of specialists who respond promptly to shocks. This paper explores the possibility of automating this process using the Reinforcement Learning method. This article discusses the development and application of an adaptive RL agent capable of managing key economic parameters in real time to stabilize the money supply in the virtual economy of the EvE online MMO game. The software package developed as a result of the research demonstrates the ability not only to maintain system stability, but also to apply various strategies for emergency response to crises, which opens prospects for the development of self-regulation mechanisms in virtual economies.

Keywords:virtual economy, reinforcement learning, Bellman equation, mathematical modeling, dynamic regulation, neural networks