On Decision Model Adaptation in Online Optimization of a Transport System

This article is about the enhancement of myopic online decision approaches for considering longer term planning goals in the management of logistic processes in a dynamically varying environment. By means of a demand peak we simulate a severe disruption of the environment of a transport system and show that a pure myopic scheduling strategy is not able to ensure an acceptable service level in such a situation. As a remedy, we propose to adapt automatically the short term decision behaviour of the used decision making algorithm. We anticipate the instantiation of a reasonable number of decision variables in a model pre-processing step in order to break the rule of selecting the least cost but also low quality decision alternatives. Within several numerical experiments we prove the applicability and suitability of our approach.