ON RELIABLE CONTROL SYSTEM DESIGNS WITH AND WITHOUT FEEDBACK RECONFIGURATIONS *

Abstract Basically, the reliability of a system is the probability that the system will perform according This paper contains an overview of a theoretic- to specifications for a given amount of time. In al framework for the design of reliable multivariable a system-theoretic context, the specification which control systems, with special emphasis on actuator a system must meet is stability; also, since, at failures and necessary actuator redundancy levels. least for most mathematical models of systems, sta-Using a linear model of the system, with Markovian bility is a long-termattribute of the system, the failure probabilities and quadratic performance in- amount of time for which the system must remain dex, an optimal stochastic control problem is posed stable is taken to be infinite. Therefore, the and solved. The solution requires the iteration of following definitions of system reliability are a set of highly coupled Riccati-like matrix difference equations; if these converge one has a reli-used in this paper: able design; if they diverge, the design is unreliable, the system and design cannot be stabil