Abstract
There is a great concern about the quality of current AI, especially DNN (Deep Neural Network), in terms of its safety and reliability. In the case of automated driving, the risk management procedures for systems containing DNNs have been verified. In this paper, we present the whole picture in a demonstrative manner. Also, we show the problems, methods for solving them, and their significance in detail. The system-level safety analysis is connected to the development of DNN model single body, and its development process. Using this method, we aim to build a system that can fix the improvement of DNN repeatedly.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Japan Science and Technology Agency (JST), Strategic Proposal for “Establishment of New Generation Software Engineering to Ensure Safety and Reliability of AI Applied Systems”
J. Sohn, et al., Search based repair of deep neural networks. https://arxiv.org/abs/1912.12463
S. Tokui, et al., in NeuRecover: regression-controlled repair of deep neural networks with training history, SANER 2022
T. Kaneko, N. Yoshioka, in Ensuring the Safety of Entire Machine Learning Systems with STAMP S&S. Learning and Analytics in Intelligent Systems (LAIS) book series (Springer, 2022)
ISO/PAS 21448:2019 Road Vehicles—Safety of the Intended Functionality, ISO Std. Jan 2019. https://www.iso.org/standard/70939.html
AI Product Quality Assurance Consortium, AI Product Quality Assurance Guidelines, http://www.qa4ai.jp/
Japan Automobile Manufacturers Association, Safety Evaluation Framework for Automatic Driving Ver1.0, http://www.jama.or.jp/safe/automated_driving/pdf/framework.pdf
National Institute of Advanced Industrial Science and Technology, Machine Learning Quality Management Guidelines, 2nd ed. https://www.digiarc.aist.go.jp/publication/aiqm/guideline-rev2.html
M. Hoss, et al., A Review of Testing Object-Based Environment Perception for Safe Automated Driving, https://arxiv.org/abs/2102.08460
R. Salay, K. Czarnecki, et al., in PURSS: Towards perceptual uncertainty aware responsibility sensitive safety with ML, SafeAI 2020.
R. Salay, et al., in A Safety Analysis Method for Perceptual Components in Automated Driving, 2019 IEEE ISSRE
S. Burton, A causal model of safety assurance for machine learning, https://arxiv.org/abs/2201.05451
IPA: STAMP/STPA (Application Edition) for Beginners—Future Safety from Systems Thinking, 2018
N.G. Leveson, Engineering a Safer World, MIT Press
T. Kaneko, et al., Ensuring the Safety of Entire Machine Learning Systems with STAMP S&S, LAIS book series, vol. 1 (Springer, 2022)
T. Kelly, R. Weaver, in The Goal Structuring Notation—A Safety Argument Notation, Proceedings of the Dependable Systems and Networks 2004
BDD100K, https://www.bdd100k.com
Acknowledgements
The authors would like to express their sincere thanks to the eAI project for their great cooperation in this research. This research was supported by the Grant-in-Aid for Scientific Research on “Establishment of Accident Analysis Methodology for Socio-technical Systems by System Theory (21K21301)” and the Japan Science and Technology Agency (JST) Project for the Creation of Future Society JPMJMI20B8.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2023 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Kaneko, T., Takahashi, Y., Yamaguchi, S., Hashimoto, J., Yoshioka, N. (2023). Safety and Risk Analysis and Evaluation Methods for DNN Systems in Automated Driving. In: Virvou, M., Saruwatari, T., Jain, L.C. (eds) Knowledge-Based Software Engineering: 2022. JCKBSE 2022. Learning and Analytics in Intelligent Systems, vol 30. Springer, Cham. https://doi.org/10.1007/978-3-031-17583-1_7
Download citation
DOI: https://doi.org/10.1007/978-3-031-17583-1_7
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-17582-4
Online ISBN: 978-3-031-17583-1
eBook Packages: Intelligent Technologies and RoboticsIntelligent Technologies and Robotics (R0)