Evaluating the effectiveness of a lighting system that alerts drivers to pedestrians they may have missed.

Maho Irita･Ayumi Nishikawa･Akifumi Yamamoto･Shiori Shimaya･Yang Han (MitsubishiElectric)

Automatic Emergency Braking (AEB) is effective in reducing pedestrian accidents, but many drivers report feeling fear when it activates. To achieve greater safety and peace of mind, our company considers it essential to reduce such fear-related near-miss experiences. Therefore, we are developing a system that alerts drivers to the presence of pedestrians before AEB is triggered, helping them become aware earlier and potentially avoid emergency braking situations.

Trajectory prediction and planning using combinatorial optimization in highway merging scenarios.

Koji Oya･Hiroshi Fujimoto (Mirise Technologies)･Yoshiaki Irie･Tomoaki Morimoto (Toyota Motor)･Kota Matsuura･Kenshin Yamamoto (Mirise Technologies)

Highway merging task requires predicting the trajectories of other vehicles and planning the trajectory of ego vehicle performing lane changes with appropriate relative speed and distance within a limited time and space, and it is one of the most challenging tasks in autonomous driving. To obtain a desirable solution from the combination of multiple vehicle trajectories, we devised the use of combinatorial optimization and conducted verification using a quantum-inspired machine.

Development of Localization Function for Achieving Autonomous Driving in Failing Environment of GNSS

Takahiro Sakai･Noriyasu Hasejima･Teppei Saitoh (Hitachi)

We developed a method to enhance localization function by integrating the results of GNSS, dead reckoning, point cloud matching, and template matching to realize autonomous driving systems in environments where GNSS is unavailable. As a result, we confirmed that autonomous driving can be sustained in GNSS-denied environments and that we can minimize tracking errors along the target trajectory.

Development of a Remote Driving Function Using a Digital Twin Environment Constructed with High-Resolution 3D Models

Noriyasu Hasejima･Kenta Maeda･Yoshibumi Fukuda･Tsuyoshi Kitamura･Hiroyuki Yamada･Naoyuki Tashiro (Hitachi)

To realize unmanned operation of autonomous driving services, the remote assist function is considered one of the key technologies as a backup for emergency situations. In this presentation, we report on the development of a remote assist function with low communication bandwidth that supports the process from stopping to resuming autonomous driving, by reflecting driving commands in a digital twin environment that reproduces the real world as a high-resolution 3D model.

Precise Docking Control of Urban Autonomous Driving using Multiple-Coordinates-based Cost Function

Hidemi Ando･Yuki Shiozawa･Takashi Fukushige (Nissan Motor)

In urban autonomous vehicles used for mobility services, it is essential to approach and stop at precise docking positions, avoiding parked vehicles and other obstacles to facilitate passenger boarding and alighting. In this study, we developed an MPC framework that combines a cost function for precise docking evaluation in Cartesian coordinates with an obstacle avoidance evaluation in Frenet coordinates. The proposed method was validated through real-world vehicle experiments.

Proposal for Stuck Vehicle Escape Control Using Tire Position Prediction in a Depression Based on Acceleration

YUJI HARA･KENTARO NISHIDA･TAKAHIRO YOKOKAWA･YOSHIYUKI IMASHIOYA (Toyota Motor)

There is a need for off-road autonomous driving to improve automobile safety. Particularly, getting unstuck requires advanced driving skills, and for those without these skills, the worst-case scenario is that they may be unable to move the vehicle at all. Therefore, this research proposes a function to get unstuck by estimating the tire's position in a depression based on acceleration and controlling the driving force according to that position.

Straight Path Tracking Control of a Trailer for Car-Caravan Type Articulated Tracked Vehicles Using Time-State Control Form

Bunji Mizukami･Yuichi Chida･Masaya Tanemura (Shinshu University)

In car-caravan type articulated tracked vehicles consisting of a tractor controlled by velocity and angular velocity and a trailer without driving force, we devised a transformation into a time-state control form along with control input conversion. This approach enabled straight path tracking of the trailer while avoiding jackknifing. The effectiveness of the proposed method was verified through simulations.