The stochastic fuel-regulator problem is investigated by optimal and self-organizing techniques. The emphasis is upon "practical realizability" of the resulting controllers. Sufficient conditions for optimality are derived. To achieve the physically realizability requirement, a specific optimal controller is introduced; and an "on-line" self-organizing control strategy is proposed that converges, in lth mean, to the specific optimal controller for a subclass of the stochastic fuel-regulator problem. The self-organization utilized is an adaptive random strategy. The formulation also considers the effect of random switching delays in the controller; and a feasibility study is included to demonstrate the practical aspects of the proposed control strategy.