Dynamic Time-Linkage Problems Revisited

摘要

Dynamic time-linkage problems (DTPs) are common types of dynamic optimization problems where "decisions that are made now ... may influence the maximum score that can be obtained in the future" [3]. This paper contributes to understanding the questions of what are the unknown characteristic of DTPs and how to characterize DTPs. Firstly, based on existing definitions we will introduce a more detailed definition to help characterize DTPs. Secondly, although it is believed that DTPs can be solved to optimality with a perfect prediction method to predict function values [3] [4], in this paper we will discuss a new class of DTPs where even with such a perfect prediction method algorithms might still be deceived and hence will not be able to get the optimal results. We will also propose a benchmark problem to study that particular type of time-linkage problems.