AUTOSAR Activities for Realizing SDV
-Overview of Software Architecture and Standardization Activities Trend-

Masahiro Goto (AUTOSAR)

I reported AUTOSAR activities for SDV in JSAE Spring Congress 2023 forcued on Vehicle Interface. I will report the progresses of AUTOSAR activities and collaborations betwee standardization organizations.

Infrastructure Sensor Installation Support System for Geofenced Autonomous Driving

Yuto Imanishi (Hitachi)･Yasuhiro Fuse (Hitachi Astemo)

To accelerate the social implementation of autonomous driving, it is important to repeat the process of technological development, environmental improvement, and evaluation of social acceptance in a short demonstration cycle. In this study, in order to speed up the integration of geofenced autonomous driving, we investigated an installation assistance system that automatically calibrates the installation posture of infrastructure sensors, which are the components of the system.

Improvement of Movement Efficiency for Autonomous Vehicle by Model Predictive Control Considering Operational Design Domain

Teppei Saitoh･Ryu Narikawa･Shinji Ishihara (Hitachi)

Automated driving control technology has been developed in order to provide mobility solutions. This research shows a real-time vehicle control technology that considers conditions on a driving route several kilometers ahead using model predictive control to improve movement efficiency with avoiding deviations from operational design domain, which is defined for the automated vehicle, due to changes in driving environment conditions.

Development of Reinforcement Learning System for Urban Driving Task

Wei-Fen Hsieh･Katsuo Semmyo･Shin Sakamoto･Masahiko Watanabe (NTT DATA Automobiligence Research Center)

We have been researching a reinforcement learning framework for autonomous driving in urban areas and reported the effectiveness of driving behavior models acquired through reinforcement learning under specific complex scenarios. This report focuses on various urban driving scenarios, and to adapt to different road users, we expanded the reward function. The results indicate the effectiveness of the proposed reinforcement learning framework.

AI-Enhanced Energy Management in Centralized E/E Architecture
-Fail-Operational Systems and Driver-Adaptive Efficiency-

Lin Li･Thomas Zipper･Martin Schlecker (AVL Software and Functions)

This presentation delves into AI-driven energy management in vehicles employing a centralized E/E architecture. The system showcases a fail-operational low- and high-voltage energy supply with three independent energy sources. Orchestrated by a central vehicle computer, it employs AI algorithms for optimal energy distribution. Zone controllers with integrated eFuses enable precise and fine-grained energy paths. A dedicated fail-operational zone guarantees the reliability of safety-critical functions, particularly x-by-wire features in advanced vehicles (L3 and above). Furthermore, the system incorporates AI-based driver adaptation, leveraging predictive analysis of the driver's usage patterns to improve the overall energy efficiency and enhance the driving range.

Environmental Magnetic Field Map Generation and Localization Method for Autonomous Vehicles on Roads

Kyoya Ishii･Keisuke Shimono･Yoshihiro Suda (The University of Tokyo)･Takayuki Ando･Hirotaka Mukumoto (Aichi Steel)

Vehicle localization is one of the key technical factors for autonomous vehicles on roads. Popular methods include GNSS and visual methods using cameras or LiDAR. However, there are conditions where each method can fail to localize. This research introduces a new method of vehicle localization using environmental magnetic field. Here, we propose a 2-dimensional magnetic field map generation method, and localization method using the generated map.