A Comprehensive Survey of Redundancy Systems with a Focus on Triple Modular Redundancy (TMR)

摘要

Despite its maturity, the field of fault-tolerant redundancy suffers from significant terminological fragmentation, where functionally equivalent methods are frequently described under disparate names across academic and industrial domains. This survey addresses this ambiguity by providing a structured and comprehensive analysis of redundancy techniques, with a primary focus on Triple Modular Redundancy (TMR). A unified taxonomy is established to classify redundancy strategies into Spatial, Temporal, and Mixed categories, alongside the introduction of a novel five-class framework for voter architectures. Key findings synthesize practical tradeoffs, contrasting high-reliability spatial TMR for safety-critical applications against resource-efficient temporal methods for constrained systems. Furthermore, the shift toward Mixed and Adaptive TMR (e.g., Approximate Triple Modular Redundancy (ATMR), X-Rel) for dynamic and error-tolerant applications, such as Artificial Intelligence (AI) acceleration, is explored. This work identifies critical research gaps, including the threat of Multi-Bit Upsets (MBUs) in sub-28nm technologies, the scarcity of public-domain data on proprietary high-integrity systems, and the absence of high-level toolchains for dynamic reconfiguration. Finally, suggestions are offered for future research directions, emphasizing the need for terminological standardization, MBU-resilient design methodologies, and the development of open-source tools for adaptive fault tolerance.