Editor's notes: GPUs have seen an increased adoption in autonomous systems. This article assesses the fault coverage that can be attained through software self-test strategies for in-field test of GPUs. - Nicola Nicolici, McMaster University@en

ISO 26262 requires classifying random hardware faults based on their effects (safe, detected, or undetected) within integrated circuits used in automobiles. In general, this classification is addressed using expert judgment and a combination of tools. However, the growth of integrated circuit complexity creates a huge fault space; hence, this form of fault classification is error prone and time consuming. Therefore, an automated and systematic approach is needed to target hardware fault classification in automotive systems on chips (SoCs), considering the application software. This work focuses on identifying safe faults: the proposed approach utilizes coverage analysis to identify candidate safe faults considering all the constraints coming from the application. Then, the behavior of the application software is modeled so that we can resort to a formal analysis tool. The proposed technique is evaluated on the AutoSoC benchmark running a cruise control application. Resorting to our approach, we could classify 20%, 11%, and 13% of all faults in the central processing unit (CPU), universal asynchronous receiver–transmitter (UART), and controller area network (CAN) as safe faults, respectively. We also show that this classification can increase the diagnostic coverage of software test libraries targeting the CPU and CAN modules by 4% to 6%, increasing the achieved testable fault coverage.@en

In order to match the strict reliability requirements mandated by regulations and standards adopted in the automotive sector, as well as other domains where safety is a major concern, the in-field testing of the most critical devices, including microcontrollers and systems on chip, is a crucial task. Since the controller area network (CAN) bus is widely used in the automotive domain, the corresponding controller ubiquitously appears in all these devices. This paper presents a generic and systematic methodology to develop an effective in-field test procedure for CAN controllers based on a functional approach (i.e., on the adoption of self-test libraries). The method can be customized to match the requirements coming from different scenarios, and allows the test engineer to maximize the achieved fault coverage in terms of structural faults in the different cases. The experimental results we gathered on a representative CAN controller model show that, given two typical testing scenarios, we are able to detect (Formula presented.) and (Formula presented.) of stuck-at faults, respectively, hence demonstrating the effectiveness of the proposed approach.@en