Indexability and Whittle Index for Restless Bandit Problems Involving Reset Processes
Source: University of California
In this paper, the authors studied a class of RMAB problems that model the monitoring and control of multiple reset processes. They showed that the RMAB satisfies Whittle indexability. When the RMAB satisfies certain monotone conditions on the stochastic evolution of the arm state, Whittle index was established in closed-form. Furthermore, when arms are stochastically identical, they showed that Whittle index policy is equivalent to the myopic policy. Based on this equivalency, the structure and optimality of Whittle index policy were established under certain conditions.