The purpose of this paper is to present a feasible strategy of adaptive control for a class of linear stochastic systems whose dynamics are not completely known. The system considered is modeled by an Itô stochastic differential equation with unknown parameters. The principal line of approach is to establish a suboptimal control by a new concept, the feasible-dynamic-programming-feedback adaptive control, which plays a joint role of learning and control. The role of learning is to generate information about the unknown system parameters, and the role of control is to realize a suboptimal control.
In order to emphasize the feasibility of the present method, an illustrative example is developed. It is shown that the present method saves computation time and computer storage and that the suboptimal control scheme obtained shows a good performance which is close to that of the optimal control. This method is compared with another method of adaptive control.