Optimal control of nonlinear stochastic systems

The paper is concerned with the development of algorithms for near-optimal control of nonlinear stochastic systems with noisy-state observations. It is assumed that the control signal acts linearly on the plant and is a function of the output. The noise processes involved in the state and observation equations are taken to be independent and white-Gaussian processes. The index of optimisation is selected to be the minimisation of the variance of the error of control. Two 2nd-order approximations to the optimal solution are derived on the basis of recently established results.