AUTOSAR Activities for Realizing SDV
-Software architecture for High Performance Computer in Vehicle-

Masahiro Goto (AUTOSAR)

In the 2nd report, the overview of software architecture of the SDV is shown with various standardization activities point of view. In this report, we discusse the configuration of in-vehicle high-performance computers and the technologies required to realise them.

Improvement of Multiple Service Metrics in Autonomous Driving Mobility by Model Predictive Control

Teppei Saitoh･Noriyasu Hasejima (Hitachi)

An autonomous driving mobility solution that can improve multiple service metrics using Model Predictive Control (MPC) is developed. This research shows a driving control method, which prioritizes the improvement of metrics expected to decline by sequentially adjusting the prediction horizon in the MPC according to the number of assigned travel destinations.

Verification of model using reinforcement learning for autonomous driving in urban area

Katsuo Semmyo･Shin Sakamoto･Masahiko Watanabe (NTT DATA Automoviligence Research Center)

Our study aims to obtain a driving model for automatic driving in urban areas with various traffic participants, such as cars and pedestrians, by reinforcement learning in an abstracted representation space. In this report, we describe our research and development efforts to apply the obtained models to a real driving environment (driving simulator).

Technology to support improvements of advanced driver assistance systems using a multimodal large language model with in-vehicle camera images

Masafumi Tsuyuki･Zhiyuan Luo (Hitachi)･Taminori Tomita･Yoshitaka Atarashi (Hitachi Astemo)

Technology that extracts desired scenes from in-vehicle camera images without visual confirmation is necessary for the efficient improvement of advanced driver assistance systems. In this study, we propose a technology that supports visual confirmation using a multimodal large language model. The evaluation results showed that the proposed method could extract half of the false positive and false negative scenes for the lane change judgment function with visual confirmation of approximately 20% of all data. This result means that the proposed method can lead to more efficient system improvements.

Measurement of Reflection and Transmission Coefficient of Snow for Camera, LiDAR, and Millimeter-wave Radar.

Yoshihisa Amano･Hiroshi Kuroda･Aiko Hibino･Masashi Mizukoshi･Hideo Inoue (Kanagawa Institute of Technology)･Kengo Sato (2309107)

Implementation of autonomous driving in 100 locations nationwide by 2027 is going. One-third to one-quarter of Japan's land will be snowy area. Our DIVP project has developed a simulator to verify the safety of autonomous driving in a virtual world rather than in the real world. In this study, we measured the reflection and transmission coefficient of snow for cameras, LiDARs, and millimeter-wave radars. For LiDAR, fresh snow was a homogeneous Lambertian reflector, and on the other hand, frozen polished "ice rink" like snow showed strong specular reflection. For reflection coefficient for radar, we found that measured curve of fresh snow and theoretical curve of relative dielectric constant of 1.4 match well. For the transmission coefficient, we found multiple reflections along thickness, which we have never seen in previous literatures, and in order to measure such extremely ne multiple reflections, we propose a unique method with a tilt snow sample.

Study of Connected Autonomous Driving in Snowfall Environment Using Optical Fiber.

Fumihito Yamaguchi (SUBARU)

In order to perform automated driving in places where GNSS signals cannot reach or to capture vehicles other than connected cars, it is important to link with information held on the road infrastructure side.
By laying optical fiber on the road, the running load of a moving vehicle can be detected and measured in terms of strain, vibration, and temperature distribution.
We have developed a cooperative automatic driving system using optical fiber sensing technology that enables real-time analysis of vehicle running conditions and road conditions, and we report on the details of the system.