Adaptive Control of Stochastic Linear Systems with Unknown Parameters.
Abstract
The thesis considers the problem of optimal control of linear discrete-time stochastic dynamical system with unknown and, possibly, stochastically varying parameters on the basis of noisy measurements. It is desired to minimize the expected value of a quadratic cost functional. Since the simultaneous estimation of the state and plant parameters is a nonlinear filtering problem, the extended Kalman filter algorithm is used. The open-loop feedback optimal control technique is investigated as a computationally feasible solution to the adaptive stochastic control problem. The open-loop feedback optimal control system adaptive gains depend on the current and future uncertainty of the parameters estimation. Thus, the standard Separation Theorem does not hold in this problem. Suboptimal control system in which Separation Theorem is arbitrarily enforced is also considered. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 1972
- Accession Number
- AD0746056
Entities
People
- Richard Tse-min Ku
Organizations
- Massachusetts Institute of Technology