CONTROL OF A RANDOM WALK WITH NOISY DELAYED INFORMATION

摘要

We consider the control of a random walk on the nonnegative integers. The controller has two actions. It makes decisions based on noisy information on the current state but on full information on previous states and actions. We establish the optimality of a threshold policy, when the threshold depends on the last action, and the noisy information. We apply the result to flow and service control problems. [References: 11]