Ergodic control of Markov chains
Mrinal K. Ghosh, Steven I. Marcus · 1990
The ergodic or long-run average cost problem, one of the classical problems in Markov decision processes, is addressed. A brief description of Borkar's convex analytic approach is presented which treats the problem with countable state space, compact action space, and unbounded cost. The ergodic control problem is related to the finite horizon and discounted cost problems.>