This paper deals with discrete-time stochastic systems composed of a large number of N interacting objects (a.k.a. agents or particles). There is a central controller whose decisions, at each stage, affect the system behavior. Each object evolves randomly among a finite set of classes, according to a transition law which depends on an unknown parameter. Such a parameter is possibly non observable and may change from stage to stage. Due to the lack of information and to the large number of agents, the control problem under study is rewritten as a game against nature according to the mean field theory; that is, we introduce a game model associated to the proportions of the objects in each class, whereas the values of the unknown parameter are now considered as "actions" selected by an opponent to the controller (the nature). Then, letting $N \to \infty $ (the mean field limit) and considering a discounted optimality criterion, the objective for the controller is to minimize the maximum cost, where the maximum is taken over all possible strategies of the nature.
4 December 2014 in International Journal of Control
In this paper, we propose an application of the so-called games against nature for giving solution to an ergodic control problem governed by a general class of Markov diffusion processes whose coefficients depend on an unknown and non-observable parameter. To this end, we assume that the values of the parameter are taken by means of ‘actions’ made by some opposite player of the controller (the nature). Then, the problem reduces to finding optimality for the controller given that the nature has chosen its best strategy. Such a control is also known as the worst case optimal control. Our analysis is based on the use of the dynamic programming technique by showing, among other facts, the existence of classical (twice differentiable) solutions of the so called Hamilton Jacobi Bellman equation. We also provide an example on economic welfare to illustrate our results.
This article concerns n-dimensional controlled diffusion processes. The main problem is to maximize a certain long-run average reward (also known as an ergodic reward) in such a way that a given long-run average cost is bounded above by a constant. Under suitable assumptions, the existence of optimal controls for such constrained control problems is a well-known fact. In this article we go a bit further and our goal is to introduce a technique to compute optimal controls. To this end, we follow the Lagrange multipliers approach. An example on a linear-quadratic system illustrates our results.