Ergodic Linear-Quadratic Optimal Control with Stochastic Coefficients

Source abstract

We investigate ergodic linear-quadratic optimal control problems with stochastic coefficients and introduce the associated stochastic ergodic Hamilton-Jacobi-Bellman (HJB) equation, in which the classical ergodic constant is replaced by a random process $λ$. We establish well-posedness of the equation and show that $λ$ characterizes the optimal ergodic costs for a general class of averaging kernels, including those corresponding to normalized finite-horizon and discounted costs. For random-periodic coefficients, we construct a random-periodic solution and show that the optimal ergodic costs associated with all admissible averaging kernels coincide. We also establish stability of the optimal ergodic costs under perturbations of the coefficients.