PROBLEMS OF SINGULARITY AND DUALITY IN LINEAR ESTIMATION AND CONTROL.
Abstract
Results in the estimation of the state of linear systems in the presence of Gauss-Markov noise processes are examined and a new proof is provided for these results. The results are then extended in several directions, namely: to continuous-time filtering of systems with measurements corrupted by noise with non-invertible covariance matrix; to optimal control problems whose cost is not a positive-definite function of the control; and to discrete-time colored-noise filtering problems. The extension to optimal control is accomplished by exploiting the duality of linear estimation and control, which is derived from duality concepts in mathematical programming. The extension to discrete-time filtering is further extended to provide a useful method of reducing the number of simultaneous equations required by the usual white-noise, discrete-time Kalman filter. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 01, 1970
- Accession Number
- AD0708480
Entities
People
- Kurt William Simon
Organizations
- University of California, Los Angeles