A Generalization of Dynamic Programming for Pareto Optimization in Dynamic Networks

Clemon University, Department of Mathematical Sciences, Clemson, South Carolina 29634-1907, U.S.A.

Received: July 1996

Abstract

The Algorithm in this paper is designed to find the shortest path in a network given time-dependent cost functions. It has the following features: it is recursive; it takes place bath in a backward dynamic programming phase and in a forward evaluation phase; it does not need a time-grid such as in Cook and Halsey and Kostreva and Wiecek's "Algorithm One”; it requires only boundedness (above and below) of the cost functions; it reduces to backward multi-objective dynamic programming if there are constant costs. This algorithm has been successfully applied to multi-stage decision problems where the costs are a function of the time when the decision is made. There are examples of further applications to tactical delay in production scheduling and to production control.