Development of Autonomous Driving with a Bird's Eye View Perception System Integrated with Map Information of Road Structure and Connectivity

Kento Kubota･Hajime Oyama･Kazuki Ishii･Kaito Sakogawa･Hideyuki Takao (SUBARU)

We developed a surrounding vehicle sensing system for autonomous driving that integrates Bird’s Eye View perception with road structure and connectivity. Surrounding vehicles detected by the perception are matched onto a map, and their future behavior are estimated by considering road structure and connectivity. This system enables autonomous vehicles to operate in coordination with surrounding vehicles based on their behavior prediction.

An Integrated Framework for Predicting Vehicle-to-Vehicle Interactions and Planning Based on GameFormer

Masaya Miwa･Akira Ito (Aichi Institute of Technology)･Ken Kinjo (DENSO)

In autonomous driving, planning faces increased state dimensionality due to interactions among multiple traffic participants, making it difficult to handle with model-based methods alone. This study proposes a planning method that integrates GameFormer, which predicts future trajectories considering interactions among traffic participants from past trajectories and map information, with model-based planning.

Trajectory planning and deviation monitoring with combinatorial optimization in highway merging scenarios

Ryuji Takahashi (MIRISE Technologies)･Koji Oya (DENSO)･Kota Matsuura･Kenshin Yamamoto (MIRISE Technologies)

We have reported a trajectory planning method by using combinatorial optimization for vehicles merging from a local road onto a highway in various situations. In the previous spring conference, we showed a comparison between simulation and real-world data in highway merging scenarios. In this report, we incorporate a monitoring function into the previous method. We show that the function can reduce the number of the planning recalculations, and hence the computational cost is reduced.

Autonomous Merging System in Traffic Congestion with Dynamic Adjustment of Decision Thresholds

Takumi Iwasa･Hanwool Woo (Kogakuin University)

This study achieves autonomous merging for automated vehicles during traffic congestion by implementing a decision-making algorithm that accurately mimics human driving behavior. Utilizing a simulation environment constructed from empirical traffic flow data, the proposed system dynamically adjusts the specific thresholds required for merging maneuvers. These adjustments are continuously made in response to the real-time behaviors of mainline vehicles, including their travel speeds and inter-vehicular distances. By employing this dynamic adjustment mechanism, the system successfully enables flexible and adaptive merging decisions under dense traffic conditions, which are inherently difficult to achieve when relying solely on conventional, fixed decision thresholds.

Analysis and Modeling of Interactive Low-Speed Merging Behavior Using a Driving Simulator

Ren Kuroyanagi･Tatsuya Ishiguro･Hiroyuki Okuda･Tatsuya Suzuki (Nagoya University)

This study analyzes and models human driver behavior during low-speed right-angle merging from a parking area onto a main road. Based on observational experiments using a driving simulator, the study focuses on a slight forward movement behavior near the stop line, referred to as “Inching” and proposes an estimation model that represents human merging decisions as transitions among multiple states.