Reliability analysis of hard real-time systems in the presence of controller malfunctions

Abstract

Hard real-time systems can fail catastrophically if their control input is not updated before the real-time deadline is reached. A faulty controller in a hard real-time system may either fail to deliver the appropriate controller outputs to the plant for one or more sampling periods or it may generate erroneous controller outputs to the plant. Both instances result in what is known as dynamic failure, that is, the system becomes unstable, which may result in catastrophic damage to the plant. For applications where high integrity or dependability is required, it is important to assess the impact of dynamic failure on the overall reliability of the system. Leveraging concepts from system control theory, we now know that it is possible for a system to survive some temporary controller malfunctions because of plant dynamics. The authors present a real-time reliability model that elaborates relationship between the fault/error processes, recovery process and the deadline of a real-time control system. To ascertain the effects of the real-time deadline of the system, they review a technique for deriving the real-time deadlines for a class of linear time invariant control systems. This deadline information is then incorporated into the real-time reliability model, and the probability of dynamic failure is calculated. Finally, they give a example of deriving the deadline for a application, and evaluating the effect dynamic failure has on the system reliability.