• Session No.123 Automated Driving and Advanced Driver Assistance IV
  • October 24Sakura Hall 29:30-11:10
  • Chair: Manabu Omae (Keio University)
No. Title・Author (Affiliation)
177

High-precision Registration Technology for LiDAR Point Clouds and Stationary Structure Point Cloud Maps to Improve Object Recognition Performance for Urban Autonomous Driving

Takashi Ikegami・Yuki Shiozawa (Nissan Motor)

In urban autonomous driving, it is necessary to accurately distinguish and recognize stationary structures such as buildings and vegetation near the road from moving objects such as vehicles and pedestrians. In this study, we developed a technology to accurately overlay LiDAR point clouds onto stationary structure point cloud maps by correcting temporal distortions in the LiDAR data and applying coordinate transformation using multiple landmarks such as poles. This technology enables reliable detection of moving objects.

178

Pedestrian's Trajectory Prediction with Fluid Dynamics for Urban Autonomous Driving

Kazuto Futawatari・Takashi Fukushige (Nissan Motor)

In order to achieve urban autonomous driving, it is necessary to make decisions on vehicle behavior near crosswalks where multiple pedestrians come and go in various directions. In this study, we investigated a method for predicting pedestrian trajectory that are essential for making decisions on vehicle behavior. The method assumes a flow field, utilizing the shape of the crosswalk and the road on the map as boundaries, to predict streamlined pedestrian trajectories from the sidewalk to the crosswalk.

179

Drivable Area Boundary Generation Method using Spatio-temporal Risk Map with Integrated Longitudinal and Lateral Control for Urban Autonomous Driving

Yuki Tanaka・Takashi Fukushige (Nissan Motor)

For autonomous driving system in urban area, it is necessary to ensure proper distance between ego vehicle and surrounding vehicles and to achieve smooth behavior. To achieve this requirement, this paper propose boundary generation method of drivable area using behavioral exploration in risk space-time map based on predicted trajectories of surrounding vehicles. This boundary is used for longitudinal MPC related to acceleration control and lateral MPC related to steering control. To confirm the effectiveness of this method, vehicle test driving is conducted in public road.

180

Stuck State Detection Technology in Remote Monitoring System for Urban Autonomous Driving

Jingze Dai・Masahide Nakamura (Nissan Motor)

In driverless autonomous driving, there is a possibility that the vehicle gets stuck due to the function or performance limitation. In this study, we propose a detection mechanism that can accurately and quickly identify the stuck state among stops in order to remotely assist in resolving the situation. The results of the experiment demonstrate that the proposal shortens the time required to resolve the stuck state and mitigates its impact on the traffic flow.

Back to Top