A fault-tolerant programmable voter for software-based N-modular redundancy

Two-Phase Low-Energy N-Modular Redundancy for Hard Real-Time Multi-Core Systems

“…Heavy fault-detection mechanisms (called HFD): with high fault-detection overheads but relatively high fault coverage. For this case we assumed that the system uses multiple fault-detection mechanisms based on code and data redundancy, arithmetic code, consistency check, and control flow checking [9], [10], [23], [24], [25], [26] to achieve high fault coverage for different fault types.…”

“…It should be noted that while both time and energy overheads of software faultdetection mechanisms are lower than the overhead of modular redundancy with result comparison (majority voting), the fault coverage of software mechanisms is not sufficiently high, unlike majority voting that provides high fault masking [9], [10], [23], [24], [26]. Furthermore, these software fault-detection mechanisms are application-specific so that each task requires its specific detection mechanism [9], [10], [25], [26], while result comparison and majority voting are general and can be used for any type of tasks without requiring any hardware modification or redesign [9], [10], [25].…”

“…8 shows, LE-TMR has a far less probability of failure than the implementations of [5], even compared to the implementation of [5] that uses heavy fault-detection mechanisms ([5]-HFD). This is because of the superiority of majority voting (NMR) in covering the faults as compared to faultdetection mechanisms [9], [10], [24], [25], [26]. 4.…”

“…An interesting observation from the experiments is that none of the implementations of [5] can achieve high reliability (the implementations of [5] cannot achieve a probability of failure less than 10 −3 ) while LE-NMR satisfies the required reliability level of safety-critical applications as they may require probability of failure be less than 10 −9 [6], [9], [10]. This is because the implementations of [5] use software fault-detection mechanisms while the fault coverage of these mechanisms is not sufficiently high [9], [10], [25], [26], unlike LE-NMR that uses majority voting that provides high fault masking [9], [10], [24], [25], [26].…”

Fault Detection in Navigation System of ALV Based on Symmetrical Behavior of Control Channels

“…Although, these methods have been successfully verified by theoretical analyses and engineering tests, all these methods are derived from accurate models that are difficult to obtain in practical engineering, especially for complicated ALVs. Besides, because dynamic models are often very complex, observer-based approach is impractical, time-consuming, requiring great process knowledge and effort [2]. Data-based analytical redundancy (soft sensor) is used whenever there is insufficient a priori knowledge about a process [2].…”