Stochastic Control for Systems with Uncertain Parameters.
ILLINOIS UNIV AT URBANA-CHAMPAIGN COORDINATED SCIENCE LAB
Pagination or Media Count:
We consider the problem of controlling a linear system with unknown random parameters and imperfect measurements. Such models are used in engineering applications, biological or economic systems containing uncertainties. This problem generally leads to dual control. We develop two approaches to the dual control problem. One is based on fixing the structures of a dynamic controller and on an approximation of the information state by the wide sense information state. The second approach is to achieve a compromise between good performance and accuracy of estimation of designing a control law which incorporates sensitivity functions as feedback signals. A fixed estimation cost is considered. Through the relationship between sensitivity and accuracy of estimation a control law is designed such that the estimation budget is rationally distributed. The two approaches give cautious controllers explicit solutions for the control law are possible and both of them have adaptive features. Simulation studies were conducted and a significant improvement over the certainty equivalent control is obtained. Author
- Statistics and Probability