Optimal Reduced-order Filter Design for Descriptor Systems with Missing Measurements

Based on singular value decomposition, a descriptor system with missing measurements is equivalently transferred to two reduced-order subsystems. Using the innovation analysis approach, the state filters are respectively given for two reduced-order subsystems with missing measurements. The computational formula for estimation error covariance matrix between two reduced-order state filters is derived. Further, the state filter for the original descriptor system is obtained. The filtering algorithm under complete measurement data in previous references has lost the optimality. A simulation example verifies the effectiveness of the proposed algorithms.