Multi-Agent Traffic Simulation and Road Safety Assessment: Current Status and Future Work

Jun Tajima (Misaki Design)･Keisuke Suzuki (Kagawa University)

This paper depicts and clarify the features of the traffic safety assessment by multi-agent traffic simulation by stating that it is essentially a numerical solution of simultaneous stochastic and nonlinear differential equation systems, and by making comparison with the results by the conventional methods such as time-series reliability model. Future issues for the advancement of the methodology will also be discussed.

Analysis of motorcycle rider risk-taking behavior based on emotional state and driver characteristics

Keisuke Suzuki･Naoki Nishiyama (Kagawa University)･Yoshitaka Mimura･Joohyeong Lee (Honda R&D)

A basic analysis was conducted on 30 participants regarding the risk-taking behavior of motorcycle riders, based on emotional states equivalent to neutral and impatient in Russell's Circumplex Emotion Model, and on driver personality traints clustered by DSQ, etc. Furthermore, a methodology for analyzing driving behavior based on these emotional states and the driver's personality traits using multi-agent simulation is outlined.

Development of Mixed-Reality Pedestrian Simulator for constructing pedestrian agent model and Proposition of Simulator Experiment Database for sharing Experimental Raw Data

Keita Oda･Jun Tajima (Misaki Design)･Keisuke Suzuki (Kagawa University)

We present the results of the observation of the interaction between the pedestrian in the Mixed Reality pedestrian simulator, which displays integrated view of real images and computer graphics and allows the pedestrian subjects walk safely, and the driver in driving simulator, which have connected each other. In addition, we discuss and propose the construction of the database to share the time-series data acquired in various simulator tests for more effective research advancement in cooperative areas.

Effectiveness Validation on Highway Merging Support System using Multi-Agent Traffic Simulation and Multiple Driving Simulator Coordination

Tohru Yoshioka (Mazda Motor Corporation, Kagawa University)･Keisuke Suzuki (Kagawa University)･Hironori Suzuki (Toyo University)･Jun Tajima (Misaki Design LLC)

Taking CACC-based support system as an example of highway merging support measures, we investigate the effectiveness of the system on merging performance and driver’s acceptability according to the penetration rate of CACC system using the virtual environment where 10 simulator vehicles can drive simultaneously in simulated traffic flow, developed as a platform to evaluate validity of various highway merging support measures. We also discuss the usefulness of the platform as an evaluation environment.

Design methods for high sense of security visibility support devices
-Utilization of multi-agent simulation Re:sim-

TAKUYA IZUMIGUCHI･KENTO MERA･HIDEKI SHINSAKA･SEISHI TAKAGI･KAN KOUNO (PENSTONE)･KOKI MIYAMOTO･TOHRU YOSHIOKA･KEISUKE SUZUKI (Kagawa University)

The effect of the visual information by the digital rear-view display on the create a sense of security was evaluated by using multi-agent driving simulator. From the results, we gained insight into the proper field of view angle, attachment position of the camera and visual information leading to predictive driving, which would help drivers to change lanes in more safe manner.

Proposal and Effectiveness Test of Electric Kickboard Metacognitive Driving Education

Rintaro Yoshikawa･Keisuke Suzuki (Kagawa University)･Kensuke Umazume (Aioi Nissay Dowa Insurance Co., Ltd.)

We proposed a metacognitive driving education program specifically designed for electric kickboards to reduce traffic accidents. In order to clarify the validity of this driving education method, we developed our own VR simulator for electric kickboards and conducted a subject experiment with 30 experimental participants. Specifically, subjects were encouraged to drive safely by using quantitative values of driving skills and a bird's-eye view that allowed them to look back at their driving objectively from a third-person perspective. As a result, we confirmed an average improvement of approximately 22% in driving ability and 14% in meta-cognitive ability.