EXISTENCE OF OPTIMAL POLICIES IN STOCHASTIC DYNAMIC
PROGRAMMING
Lawrence D. Brown
Bharat T. Doshi
Abstract: This paper deals with a general discrete-time stochastic dynamic programming
model. Under rather general conditions on the cost functions and the law of motion it is
shown that there exists a fully optimal Borel measurable policy, that is, a policy which is
optimal for future at every stage and every possible history of the process up to that stage. For
the stationary dynamic programming model this implies the existence of a fully optimal
stationary policy.
2000 AMS Mathematics Subject Classification: Primary: -; Secondary: -;
Key words and phrases: -