General method of control synthesis in a stochastic nonlinear system on the basis of a local quadratic criterion (Q1386955)

A problem concerning the control of a dynamic nonlinear stochastic system is considered. A method of synthesis of a quasi-optimal algorithm is presented. The performance index has the form of a quadratic form dependent on the state error and the control. The system disturbance is generated by Gaussian white noise. The measurement error takes also the form of Gaussian white noise. The initial state is random with given mean and covariance matrix. The components of the admissible control belong to some given closed intervals. The control which minimizes the conditional mean of the performance index, given the observation of the system output, is given. The control law where the unknown states are estimated using the Kalman filter procedure based on the linearlized equations is derived. The English translation of the original version of the paper is not the best one.