Optimal digital simulations for random linear systems with integration constraints

Abstract A generalized approach involving concepts from optimization theory is developed for realizing optimal digital simulations for linear, time-varying, continuous dynamical systems having random inputs by modifying discrete input signal variances. The minimization of a cost functional based on the state covariance matrices of the continuous system and its discrete model leads to a two-point boundary value problem which can be solved by known numerical techniques. The result is a systematic procedure for determining optimal digital simulations under the constraints that the numerical integration formula and integration step size have been specified in advance. An example is presented to illustrate the procedure, including a verification using Monte Carlo simulation runs.