A Differential Game Approach to State Estimation,
Abstract
Several approaches are formulated to the problem of estimation of the state of a dynamic system when the model contains unknown parameters. The fundamental concept can be considered as 'worst case design.' The estimator has the structure of a Kalman filter whose parameters are determined to minimize mean squared estimation error for the worst possible set of time-varying parameters. The problem is similar to a differential game in which one antagonist controls the estimator parameters and the second antagonist controls the uncertain model parameters, with conflicting goals regarding the mean squared estimation error. The problem is assumed to have a saddlepoint solution so that the cost function can be simultaneously minimized and maximized, leading, in general, to a more tractable solution. The two approaches considered involve either constraining the unknown parameter set to a compact region or penalizing the integral squared value of the parameter vector. Both approaches require computer solutions, and a basic computational approach is set up for both. It should be realized that this is merely a formulation of the problem without consideration of the existence of solutions and with no results to demonstrate feasibility or utility. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 22, 1969
- Accession Number
- AD0864177
Entities
People
- C. H. Knapp
- H. Jarvis
Organizations
- General Dynamics Electric Boat