A Time-Triggered Scheduling Algorithm for Active Diagnosis in Heterogeneous Distributed Systems

摘要

Many safety-critical control applications have now started using distributed embedded systems that consist of sensors, actuators, processors and a fully connected network with no mechanical backup. One such example is a fly-by-wire control system used in modern day aircraft. These systems require a high level of reliability, safety and performance. Many redundant components are added to ensure the reliability of such systems but this is not a viable solution because it increases the overall size of the system along with its cost and power consumption. A better approach is to continuously monitor the system to ensure that the dependability of the overall system is greater than that of its hardware and software components. Active diagnosis is one such technique that monitors the operation of the system components at runtime for fault isolation and error recovery. Since such systems are highly safety-critical i.e. their failure can result in loss of life etc., therefore it is essential that the analysis of the diagnostic information and subsequent recovery from the fault is performed within predictable time. Since scheduling the diagnostic tasks on to the system is an important aspect of this timely analysis, so the present work proposes a time-triggered static scheduler for a heterogeneous distributed system that uses diagnostic queries and a real-time database to find faults. The proposed algorithm calculates the points in time at which each diagnostic query is executed or data is replicated to or deleted from the real-time database. This a priori knowledge bounds the time for fault identification and results in a realizable diagnostic framework. The algorithm utilizes a specific priority scheme to incorporate the heterogeneity of the system and schedules the diagnostic tasks while respecting their precedence and periodicity constraints. It is prototypically implemented and is experimentally evaluated. The paper demonstrates that time-triggered scheduling can be successfully incorporated in a multi-query based diagnostic environment to increase the reliability, performance as well as to ensure the safety of a heterogeneous distributed system.