Scalable Simulation of Multiple Pedestrian and Drivers based on Behavior Models

Yuki Ban･Toru Watanabe･Hiroyuki Okuda･Tatsuya Suzuki (Nagoya University)･Takuma Yamaguchi･Chieko Nishizawa･Kazunori Ban (Toyota Technical Development)

Verification of automated driving systems requires reproduction of a wide variety of traffic conditions. The traditional model needs to be relearned based on the number of traffic participants. In this study, we will use proximity between traffic participants to build and simulate scalable pedestrian and driver behavior models.

Analysis of Decision-making of Pedestrians' Crossing Behavior for Incoming Car Fleet

Takashi Nishimoto･Hiroyuki Okuda･Tatsuya Suzuki (Nagoya University)･Chieko Nishizawa･Takuma Yamaguchi･Kazunori Ban (Toyota Technical Development)

Understanding the behavior of pedestrians, who are vulnerable road users, plays an important role in the development of intelligent vehicles. In this study, we focus on decision-making of pedestrians' crossing behavior when a line of vehicles approaches the pedestrians at unsignalized crosswalk. We report the results of observing pedestrians' behavior using a virtual environment and analyzing their decision-making of the crossing behavior.

Impact of Communication Delay on Passing Walk Behavior in a Pedestrian Simulator

Takuma Yamaguchi (Toyota Technical Development)･Hiroyuki Okuda･Tatsuya Suzuki (Nagoya University)･Eisuke Kobayashi･Kazunori Ban (Toyota Technical Development)

Remote simulations always cause communication delays, which delay the transfer of information to each other. These communication delays are an important requirement in simulator design because they have a significant impact on human interactive behavior. Therefore, we validate and evaluate the data using two simulators that can communicate with each other.

Trust Estimation of Pedestrians in Vehicles Approaching a Pedestrian Crossing in a Real Environment

Chisato Moriyama･Takuya Hamamoto･Da Li (Fukuoka University)･Toshihiro Hiraoka (JARI)･Shintaro Ono (Fukuoka University)

In this study, we conducted a confidence estimation experiment for pedestrians crossing in a real environment. Pedestrian and vehicle behaviors were captured by ground and in-vehicle cameras, and confidence was estimated by deep learning based on the distance between them and the pedestrian's skeletal coordinates. The overall correctness rates before and during crossing were 77% and 64%, respectively, revealing a trend in pedestrian behavior.