Convergence of explicit LQG self tuning controllers显式自整定控制器的收敛性.pdfVIP

Convergence of explicit LQG self-tuning controllers Prof. M.J. Grimble, MSc, PhD, DSc, CEng, MlEE Indexing terms: Stability, Phasing and phase control, Algorithms, Optimal control, Poles and zeros control law satisfied three properties. The system model Abstract: A global convergence and stability was restricted to the uncorrelated disturbance model case proof is presented for an indirect LQG self-tuning A(z-)y(t) = B(z-)u(t) + c(t), and the assumption was controller which employs a stochastic approx- made thatjhe coqmon roots of the estimated plant poly- imation type of identification algorithm. A discrete nomials (A(z -), B(z - )) remained within the unit circle linear single-input/single-output time-invariant of the z-plane as t = 00, with probability one. stochastic system with correlated noise inputs is Hersh and Zarrop [8] employed the stronger assump- considered. The plant model need not be stable or tion that (A(z-), B(z- )) remained coprime for all t T,, minimum phase. The usual assumption that almost surely. This enabled stability to be proven for the unstable common factors do not occur in the esti- pole shifting type of algorithm due to Wellstead et al. mated plant model is replaced by a weaker condi- [lo]. In each of the above contributions, the stochastic tion. The first set of stability and convergence approximation identification algorithm was used because results presented in the paper are independent of this is the simplest to analyse. the control law employed. These ar