Scholarlink

<p>As autonomous and ADAS technologies evolve, the demand for high definition (HD) lane-level maps has increased. Polyline formats, which represent lane geometry as a sequence of discrete points, are widely used in industrial HD maps due to their simplicity. However, they lack continuous geometric properties such as curvature and heading angle, making them less suitable for trajectory planning and motion control. To address this, curve-based parameterization methods have been introduced to provide a more structured representation of lane geometries. Despite their advantages, existing curve-based approaches struggle to maintain global <i>G</i><sup>1</sup> continuity and effectively model complex road structures, such as densely interconnected intersections. This study proposes an arc spline approximation framework that ensures global <i>G</i><sup>1</sup> continuity while employing a tree-based recursive parameterization approach for efficient handling of large-scale road networks. Validation using lanelet maps from the nuScenes dataset demonstrates that the framework achieves precise approximations across diverse road structures while significantly enhancing geometric fidelity compared to polyline-based formats.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Tree-Structured Recursive Arc Spline Approximation of Complex Lane-Level Maps with G1 Continuity

  • 期刊论文

  • 发表日期：

    2026年5月29日

查看全文
  • Jinhwan Jeon,
  • Seibum B. Choi

摘要

As autonomous and ADAS technologies evolve, the demand for high definition (HD) lane-level maps has increased. Polyline formats, which represent lane geometry as a sequence of discrete points, are widely used in industrial HD maps due to their simplicity. However, they lack continuous geometric properties such as curvature and heading angle, making them less suitable for trajectory planning and motion control. To address this, curve-based parameterization methods have been introduced to provide a more structured representation of lane geometries. Despite their advantages, existing curve-based approaches struggle to maintain global G1 continuity and effectively model complex road structures, such as densely interconnected intersections. This study proposes an arc spline approximation framework that ensures global G1 continuity while employing a tree-based recursive parameterization approach for efficient handling of large-scale road networks. Validation using lanelet maps from the nuScenes dataset demonstrates that the framework achieves precise approximations across diverse road structures while significantly enhancing geometric fidelity compared to polyline-based formats.