Improving CRC Fault Detection Probability in AUTOSAR E2E Based on Known Hamming Weights

Abstract

<div class="section abstract"><div class="htmlview paragraph">To develop safe vehicles, system development must be performed in compliance with functional safety. Functional safety considers situations where failures could make a vehicle unsafe, and it requires the inclusion of mechanisms to detect and mitigate these failures, even though they may not always be detected with 100% certainty — referred as diagnostic coverage (DC). Therefore, some faults, called residual faults, might go undetected.</div><div class="htmlview paragraph">In the realm of functional safety from a communication perspective, industry standards define nine distinct fault modes. The detection of these faults is crucial, especially in the widely used AUTOSAR automotive operating system. AUTOSAR E2E (End-to-End Communication Protection) serves as a communication fault detection mechanism utilizing three mechanisms: counters, timers, and Cyclic Redundancy Check (CRC) to address the nine fault modes. Especially, determining the DC for CRC can be challenging and often requires a conservative evaluation approach.</div><div class="htmlview paragraph">In conservatively evaluating the CRC fault detectability, any errors surpassing the Hamming distance threshold are inherently undetectable. However, this conservative estimation of the DC for CRC may fail to achieve the desired target residual failure probability. In such scenarios, the failure rate must be reduced or DC must be increased.</div><div class="htmlview paragraph">In this light, the present study uses the Hamming weight as a means to assess the CRC’s fault detection capabilities with a reduced degree of conservatism in comparison to that of conventional methodologies. This research also reviews the evaluation of each profile by reevaluating each AUTOSAR E2E CRC and discusses the selection of the most appropriate profile under different conditions. As a result, we improved the failure oversight rate by several orders (nearly 10 digits at maximum).</div></div>