Powell-WSC tutorial 2008

Proceedings of the 2008 Winter Simulation ConferenceS. J. Mason, R. R. Hill, L. Moench, O. Rose, eds.APPROXIMATE DYNAMIC PROGRAMMING: LESSONS FROM THE FIELDWarren B. PowellDepartment of Operations Research and Financial EngineeringPrinceton UniversityPrinceton, NJ 08544, U.S.A.ABSTRACT dimensionality: the state variable, exogenous informationand the decision variable. A major algorithmic strategy forApproximatedynamicprogrammingisemergingasapower- theseproblemsinvolvesfittingthevaluefunctionaroundthefultoolforcertainclassesofmultistagestochastic, dynamic post decisionstatevariable, whichmeasuresthestateoftheproblems that arise in operations research. It has been ap system after a decision is made but before new informationpliedtoawiderangeofproblemsspanningcomplexfinancial arrives. Thismeansthatthevaluefunctionisadeterministicmanagement problems, dynamic routing and scheduling, function of the state and action, a feature that is verymachine scheduling, energy management, health resource important in the use of scalable optimization algorithms. Inmanagement, and very large scale fleet management prob addition to this tutorial, my book on approximate dynamiclems. It offers a modeling framework that is extremely programming (Powell 2007) appeared in 2007, which isflexible, making it possible to combine the strengths of kind of ultimate tutorial, covering all these issues in farsimulation with the intelligence of optimization. Yet it re greater depth than is ...