Learning and anticipation in online dynamic optimization with evolutionary algorithms: The stochastic case

The focus of this paper is on how to design evolutionary algorithms (EAs) for solving stochastic dynamic optimization problems online, i.e. as time goes by. For a proper design, the EA must not only be capable of tracking shifting optima, it must also take into account the future consequences of the evolved decisions or actions. A previous framework describes how to build such EAs in the case of non-stochastic problems. Most real-world problems however are stochastic. In this paper we show how this framework can be extended to properly tackle stochasticity. We point out how this naturally leads to evolving strategies rather than explicit decisions. We formalize our approach in a new framework. The new framework and the various sources of problem-difficulty at hand are illustrated with a running example. We also apply our framework to inventory management problems, an important real-world application area in logistics. Our results show, as a proof of principle, the feasibility and benefits of our novel approach.