Game-theoretical and evolutionary simulation: A toolbox for complex enterprise problems

Abstract

Complex systems resist analysis and require experimenting or simulation. Many enterprise settings, for instance with cases of competition in an open market or "co-opetition" with partners, are complex and difficult to analyze, especially to accurately figure the behaviors of other companies. This paper describes an approach towards modeling a system of actors which is well suited to enterprise strategic simulation. This approach is based upon game theory and machine learning, applied to the behavior of a set of competing actors. Our intent is not to use simulation as forecasting - which is out of reach precisely because of the complexity of these problems - but rather as a tool to develop skills through what is commonly referred as "serious games", in the tradition of military wargames. Our approach, dubbed GTES, is built upon the combination of three techniques: Monte-Carlo sampling, searching for equilibriums from game theory, and local search meta-heuristics for machine learning. We illustrate this approach with "Systemic Simulation of Smart Grids", as well as a few examples drawn for the mobile telecommunication industry.