Adaptive majority voter: a novel voting algorithm for real-time fault-tolerant control systems
Gholam Reza Latif-Shabgahi, Stuart Bennett · Proceedings 25th EUROMICRO Conference. Informatics: Theory and Practice for the New Millennium · 1999
Voting algorithms have been widely used in the realisation of fault-tolerant systems. We introduce a new algorithm of software voting termed as adaptive majority voter. It uses the history of modules of a N-Modular Redundant (NMR) system to select the result of the most reliable module if it contributes toward a majority consensus. Furthermore, a new method for on-line creation of a history record of modules in a Triple Modular Redundant, TMR, system is proposed. The history vector of modules are not only used to improve the behaviour of a wide range of traditional voters, but also can be used to identify the most erroneous/faulty modules of a system to take appropriate reconfiguration or damage preventive strategies. The empirical results show that the novel 'adaptive majority voter' has higher safety and availability levels than the traditional majority voter.