Quantitative Evaluation of Safe Driving Behaviors and Their Limitations for Autonomous Vehicles in Car-to-Bicycle Crossing Scenarios
-First Report: Near-Miss-Based Risk Assessment-

Masami Aga (Tokyo University of Agriculture and Technology)･Yuichi Saito (University of Tsukuba)･Pongsathorn Raksincharoensak･Masao Nagai (Tokyo University of Agriculture and Technology)

This study analyzes 483 car-to-bicycle near miss cases and reconstructs them into collision scenarios to evaluate autonomous driving vehicle safety. The results quantitatively show that even when driving at the legal speed limit, about 43% of scenarios remain unavoidable. Preemptive deceleration can reduce collision risk; however, even with a 10 km/h reduction, 16% of cases still result in a collision, and with a 20 km/h reduction, 2% remain unavoidable. The study establishes a data driven framework for assessing safe driving behavior and clarifying both the benefits and the practical limits of autonomous driving vehicle performance.

An Evaluation of the Acceptance of a Driver Assistance System That Alerts Drivers to the Presence of Bicycles

Asuka Harada･Hitoshi Kanamori (Nagoya University)･Kenichi Yamada･Hayato Mizuma･Masaharu Saito (Toyota Motor)･Yuki Yoshihara･Nihan Karatas･Linjing Jiang･Takahiro Tanaka (Nagoya University)

Communication-based driver assistance systems, particularly V2X-based technology, are expected to prevent crossing collisions on residential roads. However, challenges remain regarding the accurately of notifications alerting drivers to the presence of bicycles. We identified the issues associated with low correct notification frequencies and narrowed down the notification methods suitable for low-frequency notifications. We conducted driving simulator (DS) experiments under various conditions, including correct notifications, false notifications, and missed notifications, to assess the acceptability of the assistance system. We also investigated the impact of low-frequency assistance systems on driver attention and report our findings here.

Effects of Viewpoint Position on Steering Characteristics in Remote Operation of Autonomous Buses

Yusaku Matsumura･Naoki Furugohri･Toshiyuki Sugimachi･Toshiaki Sakurai･Shuichi Yahagi (Tokyo City University)･Masaaki Onuki･Kimihiko Nakano (The University of Tokyo)･Jongseong Gwak (Takushoku University)

This study evaluated the impact of different viewpoint positions on driving characteristics in the remote operation of autonomous buses. Using a driving simulator for heavy-duty vehicles, we established various viewpoints and speed conditions. By conducting a comparative evaluation of driving performance and workload, we examined the optimal viewpoint conditions for remote operation.

Online Map Generation Using Stereo Cameras: A Case Study

Ryo Yanase･Langxing Tan･Akitaka Okou (SUBARU)

Accurate map generation is a key challenge in autonomous driving and advanced driver assistance systems (ADAS). This paper proposes a method for online map generation using object recognition results from stereo cameras. The proposed framework integrates observed object information in real time, supporting robust map generation that can adapt to dynamic changes in the environment. In addition, we validate the effectiveness and practicality of the proposed method through experiments using real-world driving data.