Dynamic and stochastic planning problems with online decision making: a novel class of models

In this thesis we study planning problems in the area of routing and scheduling by means of mathematical models and (numerical) optimization. Characteristic for this class of problems is that here-and-now decisions have to be made while only probabilistic information is available on the relevant planning period. Moreover, during the planning period online decisions on arriving service requests are required. A novel class of models is developed in which ideas from the research fields stochastic programming and online optimization are integrated. These models comprise two stages. In the first stage planning decisions are modeled, while the second-stage problem consists of an online optimization problem. The expected future costs of the second stage are included in the first-stage objective. For two practical planning problems models have been developed and analyzed. Numerical experiments have been conducted on semi-realistic instances.