Towards the fault tolerant software: fuzzy extension of crisp equivalence voters

Redundancy, on which fault tolerance is based, can be achieved through hardware, software, information and time. With respect to different version outputs from redundant software versions, voting strategies are separated into two classes. Voting strategies are either based on output classification, partitioning of the outputs, or on convergence functions. The traditional equivalence relation does not enable gradual comparisons below the fixed threshold. Fuzzy extension of classical numerical equivalence relation, proposed in the paper, overcomes those potential problems. Test examples are graphically illustrated