A Study on Control Calibration Optimization in THS Engine Start-up Vibration by Applying MBSE

Soh Shimizu･Masaki Sagawa･ ･ ･ ･ (Toyota Motor)

In recent vehicle dynamic performance development, increasing complexity has led to significant growth in control calibration effort and rework. This study targets engine start-up vibration in hybrid vehicles and constructs a SysML-based model. By visualizing relationships among requirements and design elements, and introducing validation based on intermediate state variables, a principled calibration approach is achieved. The proposed process demonstrates effectiveness in reducing rework and calibration effort while supporting an efficient development flow from performance planning to hardware validation.

Improving e-Axle System Development Efficiency through MBSE and Generative AI (LLM)
-(Part II)-

Takurou Hirano･Masaru Katsuki･Kazunori Kawashima･Takurou Kawasumi (JATCO)･Daiki Katagiri･Makoto Morinaga･Motokazu Nishimura (Stockmark)

In our previous report, we presented the efficiency improvement of e-Axle development by converting MBSE models into knowledge graphs and utilizing generative AI. However, a remaining challenge in maintaining the knowledge graph is the need for continuous updates. In this presentation, we propose a methodology and provide practical examples of continuously updating the graph by using AI to automatically extract information structures from unstructured documents, such as technical reports.

Investigation of Risk Assessment Methodology for Physical Layer Attacks to Prevent Vehicle Theft

Maki Shimokawa･Takanori Miyoshi･Kazuhiro Kakito･Satoshi Horihata (Sumitomo Electric Industries)

ISO/SAE 21434 risk assessments tend to underestimate physical layer attack risks. To address growing physical threats like vehicle theft, this paper proposes a risk assessment methodology for automobiles that explicitly incorporates physical-layer considerations. Furthermore, we discuss the importance of physical-layer security alongside software countermeasures to achieve early attack detection.

Bayesian Approach for Unknown Risk Quantification on Highly Automated Driving

Masanori Yokota (NTT Data Automobiligence Research Center)

In highly automated driving, it is difficult to identify all contexts and triggering conditions in advance. Therefore, in safety arguments, it is necessary to appropriately estimate the risks arising from unknown contexts and triggering conditions and evaluate whether they meet acceptable levels. This presentation will introduce a method for evaluating unknown risks using a Bayesian approach, through concrete examples.

A Re-certification Credit Allocation Method for OTA Software Updates in Software-Defined Vehicles

Naoki Hashimoto (Monstarlab)

Over-the-air (OTA) software updates can alter type-approval-relevant functions of software-defined vehicles, yet how much of the required re-assessment can be discharged virtually rather than physically remains under-specified. Simulation-credibility provisions for type approval (EU 2022/1426; NATM) and software-update management (UN-R156/SUMS, RXSWIN) exist separately, but no method connects them by accounting for virtual evidence reuse across an OTA change. This study proposes such a credit-allocation method, pairing change-impact classification with a virtual-evidence credibility tier, and demonstrates it on the OTA-delivered ODD speed extension of a deployed L3 ALKS (UN-R157).

Regulatory Frameworks for RXSWIN and Software Update, A Comparative Overview between Japan and EU

Megumi Enomoto･Tetsuya Niikuni (NALTEC)

Post-production vehicles can have their original functions extended or modified by software updates following the establishment of UN-R156 in 2021. Methods for verifying regulatory compliance of those modified vehicles are different between Japan and the EU; the EU approach specifically compares the RXSWIN of the vehicle and its certification. This study comprehensively outlines the contents and the background of the 2025 UN-R156-01 amendments, and provides an analysis of operational challenges of software updates using RXSWIN.