Accession Number:

AD0722347

Title:

Some Results in Dynamic Programming

Descriptive Note:

Research rept.

Corporate Author:

CALIFORNIA UNIV BERKELEY OPERATIONS RESEARCH CENTER

Personal Author(s):

Report Date:

1971-04-01

Pagination or Media Count:

19.0

Abstract:

The first part of the report discusses a dynamic programming model in which all rewards obtained by the decision maker are assumed nonnegative. The decision makers objective is to successively choose actions so as to maximize his expected reward earned over an infinite time span. It follows from known results that the decision makers choice need only depend upon the outcome of a randomization that depends on the model only through the state of the model and the time when the choice is made. It is shown by counterexample that this is basically the smallest class of decision rules that need be considered. Conditions under which a stationary policy is optimal are also presented. The second part of the report discusses the same model under a new criteria, namely, the average cost incurred per unit time. An example is presented in which there does not exist an epsilon-optimal randomized stationary policy.

Subject Categories:

  • Operations Research

Distribution Statement:

APPROVED FOR PUBLIC RELEASE