Computing efficient steady state policies for deterministic dynamic programs. I (Q1191780)

The author discusses efficient steady state policies (SSP) for deterministic dynamic programs in which the objective is to optimize the total undiscounted costs over an infinite or long finite planning horizon. The opportunity cost is used to measure the adverse effect of a suboptimal policy \(\pi\) oin the total cost under certain conditions discussed by the author elsewhere [SIAM J. Appl. Math. 37, 128-147 (1979; Zbl 0448.49013)]. Such an SSP is determined by solving a mathematical program to find the target state \(s^*\) and also a feasible guidance rule is indicated that moves the system to \(s^*\) from some set of states. The author claims that using it as a first-order approximation can produce good computational savings but such an SSP does not optimize transient costs. For this purpose the author also studies second-order improvements that reduce the transient costs of a given first-order approximation \(\pi\).