An Item Response Theory-Based Classification of Hazards in a Driving Risk Assessment Test

Kazumitsu Shinohara･Naoko Masuda (The University of Osaka)･Kan Shimazaki (Kindai University)･Rin Itoh･Hirofumi Aoki (Nagoya University)

The driving risk assessment test evaluated in this study requires participants to observe computer graphics (CG) first-person driving videos and sequentially touch detected hazards within each scene. To improve evaluation accuracy, it is desirable to assess hazard perception skills while accounting for the unique characteristics of each individual hazard presented during the test. This study aims to clarify the factor structure underlying hazard perception abilities within this assessment framework and attempts to estimate the item characteristic parameters of each hazard utilizing item response theory (IRT).

Characteristic Analysis of Vehicle-VRUs Crashes Based on Real Accident Data in China

ze yao Li (Huaqiao University)･Zhuo Li･Di Pan (Xiamen University of Technology)･Koji Mizuno (Nagoya University)･Yong Han (Xiamen University of Technology)

Vulnerable Road Users (VRUs) suffer high casualty rates in traffic crashes and are a major focus of road safety research. Based on 4,060 real vehicle-VRU crash videos from the VRU-TRAVi database in China, this study conducts statistical and characteristic analysis across pre-crash, in-crash, and post-crash stages. K-modes clustering is used to identify typical accident scenarios. The findings reveal spatiotemporal, kinematic, and injury patterns of pedestrians, cyclists, and powered two-wheelers, and clarify key risk factors including visual obstructions, speed, and impact locations. Results provide reliable data support for vehicle safety design, road management, and VRU crash prevention.

Study on Strength Prediction of 6000 Series Aluminum Alloys Using Machine Learning and its Application to Crash Simulation (Second Report)

Junya Nagai･Kentarou Aono･Ryousuke Negawa･Hiroki Takami (SUBARU)

The purpose of this research is to use machine learning to predict the strength of an actual vehicle after 6000 series aluminum alloy, which has age-hardening characteristics, has hardened due to heat during the paint baking process, and to apply this to a crash simulation to improve the accuracy of performance predictions by making the model more accurate to the actual vehicle.

Theoretical Evaluation of μ Measurement with Icy Road using Backscattered Light

Tomoki Kawahara･Akihiro Kido (Tohoku Gakuin University)

In the measurement of the friction coefficient (μ) of icy road surfaces using NIR backscattered images, multilayer reflections composed of reflections from both the ice surface and the underlying road surface layer are measured. In particular, the reflected light from the road surface layer is considered to depend on both the measurement wavelength and the surface roughness. In this study, these reflection characteristics were investigated, and the validity of measuring the friction coefficient (μ) of icy road surfaces using backscattered light was theoretically verified.

Modeling and Reproducibility Verification of Vehicle Rollover Behavior Induced by Entry into Gravel

Hiroki Ihara (SUBARU)･Toshihiko Kozai･Tsutomu Iwase (Gunma University/SUBARU)･Naoki Morotomi (SUBARU Techno)･Hirofumi Kinoshita･Seiichi Gokurakuji (SUBARU)

In real-world traffic environments, vehicles may skid into roadside sand or gravel, potentially leading to rollover accidents. Focusing on this gravel entry, we modeled the gravel and identified the friction coefficient between gravel particles. Furthermore, we verified the reproducibility of vehicle behavior using the developed gravel model. This paper reports on these results.

Remote Connection Demonstration for Cycling and Driving Simulators and the Impact of Latency on Decision-Making

Takuma Yamaguchi･Masayasu Harimoto･Kazunori Ban (Toyota Technical Development)

Since traffic behavior involves coordinating with others to move smoothly, linking multiple traffic simulators is a useful method for analyzing traffic behavior. However, such linking requires network communication, making latency a significant challenge. Therefore, we will demonstrate remote connections and verify the resulting latency. Furthermore, by utilizing a decision-making model, we will assess the acceptable level of latency.