Transformer-Based Assessment of Driving Corridors for Motion Planning of Automated Vehicles

Determining suitable driving corridors in arbitrary traffic scenarios is a key challenge for motion planning of automated vehicles. Driving corridors can be efficiently computed via reachability analysis and serve as solution spaces for motion planning. However, selecting a suitable driving corridor for trajectory planning is an open research question. To address this research gap, we introduce a novel, transformer-based architecture which is capable of assessing multiple set-based driving corridors in terms of 1) their drivability and 2) expected costs of the to-be-computed trajectory within them. Besides these assessments, an initial trajectory is provided as well. Our numerical experiments on CommonRoad scenarios demonstrate that our approach outperforms an established heuristic.     «

Determining suitable driving corridors in arbitrary traffic scenarios is a key challenge for motion planning of automated vehicles. Driving corridors can be efficiently computed via reachability analysis and serve as solution spaces for motion planning. However, selecting a suitable driving corridor for trajectory planning is an open research question. To address this research gap, we introduce a novel, transformer-based architecture which is capable of assessing multiple set-based driving corri...     »