CONTINUOUS CONTROL WITH STOCHASTIC SHOPPING TIME,
Abstract
The criterion for a continuous control process whose stopping time is a random variable of known cumulative probability distribution is expressed as a single improper integral. (Physical examples of interrupted control situations are: regulator component failure, catalyst deterioration, space vehicle midcourse guidance, and random waiting for a digital computer used in the control of several systems.) A necessary condition for an optimal trajectory is derived, shown to be time-invariant only for the exponential probability law, and related to reliability theory. The optimal feedback control law, optimal trajectory, and minimum expected cost are presented for time-invariant linear systems with quadratic cost and exponentially distributed stopping time. A simple condition on the feedback rule for time-varying linear systems with arbitrary stopping time probability distribution is derived. A connection with the deterministic theory is established via the limit as the variance approaches zero. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1966
- Accession Number
- AD0634831
Entities
People
- Allen Klinger
Organizations
- RAND Corporation