Study on Virtual Test Platform by utilizing Multi-agent Traffic Simulation for Sophisticated Automated Driving System Development

Sou Kitajima･Shun Endo･Nobuyuki Uchida･Kunio Yamazaki (JARI)･Naoki Suganuma (Kanazawa University)･Tadashi Okuno (OS Planning)･Jun Tajima (Misaki Design)

This paper describes an effective virtual test platform to develop more sophisticated Automated Driving System by generating both complicated and perplexing scenarios. Because it is difficult to find and accumulate these scenarios through field operational test, an enhanced function is developed and implemented our multi agent traffic simulation. The platform enables not only to provide significant variation of extracted rare scenarios but also to realize decision making improvement of Automated Driving System by efficient training for proficient interaction with other traffic participants.

Unified Path Planning Algorithm for Traffic Participants in Multi-Agent-based
Simulation and Its Parameter Estimation from Actual Data

Hironori Suzuki (Toyo University)･Jun Tajima (Misaki Design)

We propose a discrete-choice-based path planning algorithm that can be uniformly applied to vehicles, motorcycles, pedestrians, bicycles, and other traffic participants appears in the multi-agent traffic simulation. We also describe an estimation method of parameters in the path planning algorithm. The effectiveness of the proposed scheme is verified using the behavior data obtained from a simulator.

Multi-Agent Traffic Simulation based on Dashboard Camera Images
-Development of 3D Position Reconstruction Technology utilizing Digital City Models-

Toshiya Arakawa (Nippon Institute of Technology)･Jun Tajima (Misaki Design LLC.)･Hironori Niwa (nitro LLC.)

The technology to reproduce recorded traffic situation by multi-agent simulation based on the video images of Dashboard Camera would be becoming increasingly important in various application areas. This paper describes the significance of this technology and proposes a method to improve the accuracy of estimation of position, orientation etc. of traffic participants by utilizing a matching technique of the images, between the one created from digital city model and the other recorded Dashboard Camera image.

Acquisition and analysis of simulator driving data in the naturalistic traffic flow generated by multi-agent traffic simulation by the same subject over months

Toshiaki Kimura (Kyoto Tachibana University)･Hironori Suzuki (Toyo University)･Jun Tajima (Setouchi Simulator)

We present the results of an investigation of changes in driving behavior, subjective evaluation, and unconscious attitude to the traffic safety obtained from the simulator experiment in which the same subject drives the driving simulator for several months in naturalistic traffic flow generated by multi-agent traffic simulation. Also the comparative discussion is made with conventional condition-controlled and short-duration simulator testing methods and proposed naturalistic and long-duration simulator testing method shown in this paper.

Analysis of Factors Leading to "Abandoning Cars" During Disaster Evacuations

Haruto Masubuchi (Nippon Institute of Technology)･Kousuke Aoyama (formerly: Nippon Institute of Technology)･Toshiya Arakawa (Nippon Institute of Technology)

During the Great East Japan Earthquake, traffic congestion caused by vehicle evacuations led to situations where evacuees were caught in the tsunami. This study conducted simulator-based experiments to analyze factors influencing vehicle abandonment during tsunami evacuations. A tendency to abandon vehicles in conformity with preceding vehicles was observed. Additionally, it was shown that displaying the tsunami arrival time was not a direct factor influencing vehicle abandonment.

Managing automaker's legal risks in traffic accidents of autonomous driving,using Multi-Agent Traffic Simulation

Naohiro Tomochika (Mominoki Tomochika Law Firm)

When automakers assign values ​​to the programs of autonomous vehicles, they need to interpret the technical standards for autonomous driving devices. In order to manage legal risks in traffic accidents caused by their autonomous vehicle, they must be able to prove the legitimacy of the values and the legitimacy of the interpretation process. This paper considers interpretation methods that meet these demands.