Managing Operations in Chaotic Environments with Evolutionary Software Agents

Abstract

Managing dynamic systems that interact with unstable environments is crucial in various fields, including stabilizing turbulent flows in gas and hydrodynamics, handling chemical and technological processes, generating signals in radio engineering, and managing assets in capital markets. The challenge lies in the limited predictability of deterministic chaos models, which results in additional uncertainty arising from random observations, as explained by traditional statistical models. Numerical methods provide the only viable approach to evaluating management effectiveness under these complex conditions. This study focuses on empirical algorithms for identifying and predicting local trends, with the goal of developing extrapolation prediction techniques. The primary case examined in this paper is speculative trading in currency markets, known for their stochastic chaos. Unlike physical and technical problems, the currency market is purely informational and lacks inertia. As a result, traditional prediction algorithms used in software robots based on reactive control strategies have proven ineffective. This study aims to address this efficiency issue by exploring control strategies that sequentially optimize evolutionary parameters and approximate the model structure of the observation series.