ALGORITHMS FOR STATE BOUNDING IN LARGE-SCALE SYSTEMS

A class of linear and time-invariant large-scale systems under unknown input signals and measurement errors is considered. No assumptions about statistical properties of the unknown quantities are made. The uncertainty is modelled through bounds that define ellipsoids in which lie unknown initial conditions and unknown ellipsoidal tubes containing signal trajectories. The state estimation problem consists of determining on-line the smallest sets in the state space in which the unknown system state lies. Two hierarchical estimation algorithms are proposed and compared, namely the 'completely decentralized with interaction measurements' (CDIM) algorithm and the two-level hierarchical with interaction measurements (THIM) algorithm. The THIM algorithm leads to smaller estimating sets owing to additional co-ordinator-level information.