Supporting Synthetic Data-Driven Diagnosis through Automated Fault-Injection - Robotics Institute Carnegie Mellon University

Supporting Synthetic Data-Driven Diagnosis through Automated Fault-Injection

Patrick E. Lanigan, Priya Narasimhan, and Thomas E. Fuhrman
Conference Paper, Proceedings of 4th International Conference on Dependability (DEPEND '11), pp. 65 - 71, August, 2011

Abstract

Given the lack of empirical data available from automotive serial-communication networks, an automated faultinjection environment can be used to create synthetic datasets for training and testing data-driven diagnosis algorithms. We use commercial fault-injection hardware with custom software to implement such an environment. A small pilot study using injected physical-layer faults shows promise in producing identifiable error-patterns.

BibTeX

@conference{Lanigan-2011-126399,
author = {Patrick E. Lanigan and Priya Narasimhan and Thomas E. Fuhrman},
title = {Supporting Synthetic Data-Driven Diagnosis through Automated Fault-Injection},
booktitle = {Proceedings of 4th International Conference on Dependability (DEPEND '11)},
year = {2011},
month = {August},
pages = {65 - 71},
}