Improved Algorithms for Estimation, Prediction and Control.
Abstract
This report describes the results of a study to examine the feasibility of developing fast algorithms for estimation, prediction and control. The objective of the study was to assess the likelihood fo finding procedures which could be faster, in terms of computer running time, than the classifcal least-squares method of parameter estimation, used extensively to develop models of stochastic phenomena. While the least-squares method has proved its worth in over a century of use, it has some serious drawbacks, stemming from the fact that it requires the inversion of matrices. for 'large' problems such as the problem of tracking many missiles or processing data from multiple intelligence sensors in real-time, the computational burden of the least-squares method can overwhelm even today's powerful computing systems. Previous attempts to solve this problem have centered on the development of faster computers (e.g., array processors), the improvement of algorithm for matrix inversion, or the simplification of the model to produce a matrix that is easier to invert. In general, these approaches have not been successful in solving the problem. Modern sensor exploitation systems, for example, Still cannot operate in real-time or even near-real-time.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 26, 1986
- Accession Number
- ADA167601
Entities
People
- J. G. Caldwell