h-NESDF: An Infinitely Differentiable Signed Distance Between Two Sets with Robotic Applications
Computing distances between sets is important in many robotic applications, such as motion planning and control. In many applications, derivatives of the distance are needed as well.
- 用途
- 原典で確認
- 難易度の目安
- Hard
- 実装コストの目安
- Medium
概要
Computing distances between sets is important in many robotic applications, such as motion planning and control. In many applications, derivatives of the distance are needed as well. It is well known that the Euclidean distance is not always differentiable, which create issues to algorithms that need a derivative. In this paper, we introduce NESDF, the Normal Euclidean Signed Distance Function between two general (with mild regularity assumptions but possibly non-convex) sets. This metric is still not differentiable, but it allows us to introduce a generalization, $h$-NESDF, which has a smoothing parameter $h$ such that the metric is infinitely differentiable for all $h>0$ and that reduces to the NESDF when $h$ goes to $0$. The metric is amenable to GPU parallelization, and thus can be computed efficiently. We show extensive simulation results to evaluate its viability in motion planning.
参考リンク
リンク先のライセンス・再現性・データ条件をご確認ください。自動分類や実装の目安は保証ではありません。
実装を始めるプロンプト
参考リンクを確認したうえで、次の内容をコピーして利用できます。
# 論文・記事 h-NESDF: An Infinitely Differentiable Signed Distance Between Two Sets with Robotic Applications # 概要 Computing distances between sets is important in many robotic applications, such as motion planning and control. In many applications, derivatives of the distance are needed as well. # 原典 https://arxiv.org/abs/2611.03944 上記の手法を実装してください。実行前に依存関係とライセンスを確認してください。