Data-based Closed-loop System Simulation

This paper provides a framework for control design and simulation of a closed-loop system with partial input/output data of a plant. Given the input/output crosscorrelation and output autocorrelation data of an open loop dynamic system, a simulation model implemented in fixed-point digital devices which matches these data is obtained using q-Markov Covariance Equivalent Realizations. These results allow the design of digital simulations with no error within the specified set of crosscorrelation and autocorrelation data. When a linear approximation of the plant is assumed, an LQG controller can be presented solely in terms of the input/output crosscorrelation data. This is the so-called the Markov data-based LQG control. With both the simulation model of the plant and the controller in hand, a closed-loop simulation can be constructed. This is yet another example showing that significant work can be done with very limited information of a plant. Copyright c ©2005 IFAC.