policy for restless multi-armed bandits

A policy for restless multi-armed bandits is a decision rule that dynamically selects which arms to activate over time when each arm’s state evolves continuously regardless of being played, aiming to optimize a long-term performance criterion under these non-stationary conditions.