Answers are grounded in this paper and cite their sources.
Industrial embedded systems operate in safety-critical domains including aerospace, automotive, rail transit, and medical devices, where faults can cause catastrophic failure or loss of life. Under demanding environments and strict functional safety standards, achieving high reliability remains a fundamental challenge. This paper presents a systematic review of fault-tolerance and fault-handling techniques for such systems. We begin by clarifying reliability requirements and distinguishing systematic faults from random hardware faults. Next, we analyze three redundancy mechanisms: hardware, software, and time. We then survey fault detection and diagnosis methods, including watchdogs, consistency checks, EDAC, control-flow monitoring, model-based observers, signal feature extraction, and AI-driven diagnosis. A graded recovery strategy follows: fault masking, failover, system degradation, and safe shutdown—which constitutes a defense-in-depth architecture. Finally, two validated case studies are presented: a triplex flight control computer (avionics) and an ISO 26262-compliant automotive ECU. The review shows that modern fault tolerance relies not on isolated techniques, but on coordinated, multi-layered engineering. Future trends, including AI-driven predictive diagnosis, adaptive redundancy, formal verification, and RISC-V-native fault-resilient hardware, point toward more robust, verifiable, and efficient future systems.
No takes yet. Share an insight, caveat, or question.
Lanxin Zhu (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: