Active Monitoring for Control Systems under Anticipatory Semantics

Abstract

As the increment of software complexity, traditional software analysis, verification and testing techniques can not fully guarantee the faultlessness of deployed systems. Therefore, runtime verification has been developed to continuously monitor the running system. Typically, runtime verification can detect property violations but cannot predict them, and consequently cannot prevent the failures from occurring. To remedy this weakness, active monitoring is proposed in this paper. Its purpose is not repairing the faults after failures have occurred, but predicting the possible faults in advance and triggering the necessary steering actions to prevent the software from violating the property. Anticipatory semantics of linear temporal logic is adopted in monitor construction here, and the information of system model is used for successful steering and prevention. The prediction and prevention will form a closed-loop feedback based on control theory. The approach can be regarded as an effective complement of traditional testing and verification techniques.