Advanced Search
Article Contents
Article Contents

Robust multiobjective dynamic programming: Minimax envelopes for efficient decisionmaking under scenario uncertainty

Abstract Related Papers Cited by
  • Many sequential decision problems are characterized by multiple objectives and can be formulated as multiobjective dynamic programs. A subset of these problems concerns systems that are only partially observable such that the system response to implemented policies is known to belong to a set of possible system responses but is not uniquely known prior to policy selection. A new methodology is developed to identify optimal strategies in finite-horizon multiobjective decision problems for systems of this type. These strategies will either be minimax efficient with respect to a partial ordering in the multiobjective space or, where minimax efficient strategies do not exist, minimax optimal with respect to a total ordering in a scalar space induced by decisionmaker preferences over the set of objectives. In formulating the dynamic program, system uncertainty is described by a finite set of scenario-specified system models, and the likelihood of any particular scenario is assumed to be unknown. By accounting for different scenarios, the multiobjective dynamic program and the resulting strategies are robust with respect to uncertainty in the underlying policy-response relationships. The solution concept is developed with assurances that the principle of optimality holds. An illustrative example demonstrates the methodology.
    Mathematics Subject Classification: Primary: 90C29, 90C47, 90C39; Secondary: 49K35.


    \begin{equation} \\ \end{equation}
  • 加载中

Article Metrics

HTML views() PDF downloads(81) Cited by(0)

Access History

Other Articles By Authors



    DownLoad:  Full-Size Img  PowerPoint