A reduced order model-partitioned system identifier for a class of systems with input uncertainty

An adaptive algorithm, that has the capability of identifying a system with input uncertainty (fixed and unknown, but restricted to a known set of possible inputs or randomly changing in time within a known set of possible inputs) is developed, and simulation results are reported. A model-partitioned recursive least squares methodology is adopted within a Bayesian framework. Thus, a low-order identifier, which not only matches the input-output characteristics but also points out which input is going through the actual plant, is obtained