Study of Robust Design to Ensure the Manufacturing Quality of High-Density Mounted Electronic Components on the printed board.

Hisao Nishimori･Jun Muto･Tomoyuki Furukawa･Miiyu Orinaka･Yasufumi Shibata (Toyota Motor)

With the increasing demands for higher functionality in electronic components, it is essential to develop design techniques that ensure stable solder joints during the manufacturing process when implementing advanced semiconductors and components on printed circuit boards at high densities. Since lift-off of the solder joints was observed on experimental boards, this paper introduces considerations and discussions on key design parameters for robust design utilizing simulations and AI.

Predicting Engine Physical Sensor Values Using Multiple Regression Models

Yuki Yano･Kenichi Morizane･Koshiro Wada･Nobuo Yunoki･Kenta Kobayashi (Mazda)

Replacing onboard sensors with AI models offers numerous advantages, including weight and cost reduction, as well as greater flexibility in layout design. In particular, multiple linear regression models are lightweight, easy to interpret, and well-suited for implementation.
This study introduces the development process of a multiple linear regression model for accurately predicting diesel engine exhaust temperature and exhaust pressure, along with the results of validation using an actual vehicle.

Study on Improving the Robustness of Stroke Sensors for Steer-by-Wire EPS

Yohei Shirakawa･Yoshiaki Yanagisawa･Yukio Ikeda (PROTERIAL)

We are investigating an inductive stroke sensor for detecting the displacement of the rack shaft in a steer-by-wire EPS. We examined the mechanism of detection accuracy deterioration caused by the inclination of the rack shaft relative to the sensor. To mitigate this accuracy deterioration, we devised two novel sensor structures with high robustness. The effectiveness of these structures was validated through electromagnetic field simulations.

Actual vehicle electromagnetic noise evaluation and prediction technology for electrical components

Yukihiro Serizawa (Sohwa & Sophia Technologies)

As electrification progresses, EMI issues must be overcome to ensure efficient development. This technology can predict electromagnetic noise before measuring actual equipment, and by utilizing CAE, it realizes test stress and contributes to MBD. Here, we will introduce case studies and suggest ways to proceed in the future.

Development of an artificial weather chamber that reproduces a dynamic weather environment for autonomous driving sensors　(3rd report)

Satoshi Akaike･Hiroyuki Enoki･Yuri Saito･Hisayasu Shima (ESPEC)

With the advancement of autonomous driving technology, the onboard sensors may be affected by various weather conditions, which necessitates the development of new evaluation methods.
In this report, we continue from the previous study to evaluate sensor visibility under fog and rain conditions, aiming to address the upcoming mandatory installation of automatic braking systems and the enhanced performance evaluation under adverse weather conditions.

How AI shapes smooth vehicle user experience with Small Language Models - today

Dr. Johannes Richenhagen･Tobias Schafer･Dr. Dirk Macke･Jorg Kottig･Alexander Kugler･Thomas Hulshorst (FEV.io)

Generative AI with Large Language Models (LLM) faces challenges regarding latency, privacy, cost, and system and user customization. Small Language Models (SLM) address challenges of LLM with onboard privacy, real-time performance, and vehicle integration, enhancing user interaction.
This technology is demonstrated with an intuitive, user-configurable dashboard on a Software-Defined Vehicle platform, combining SLM and LLM. Spoken language is converted to text, interpreted, formalized and executed. By this, both language model benefits are leveraged for seamless driver companionship.
This enables quick roll-out of AI features at optimum costs, latency, privacy and reliability - paving the road for more AI in Software-Defined Vehicles.

AI-based low pressure fuel pump pressure sensor prediction model by Super TML

TACKOON KIM･JUNYOUNG SHIN･SOOIN LEE (Hyundai Motor)

With the development of on-board AI, we need to think about how to use integrated AI sensors in vehicles. In this paper, we verified whether AI sensors can replace hardware sensors by utilizing sensor information provided to users on behalf of sensor values of low-pressure pumps. We also verified the accuracy of AI sensors by comparing them with actual vehicle driving data. A statistical guide on which sensor values should be used as inputs was also verified.