arXiv 2511.205932026-02-02cs.ROcs.SYeess.SY

用于机器人运动规划的安全稳定神经网络动力学系统

Safe and Stable Neural Network Dynamical Systems for Robot Motion Planning

Indian Institute of Technology Kharagpur（印度理工学院科拉格浦分校）；Munich Institute of Robotics and Machine Intelligence, Technical University of Munich（慕尼黑机器人与智能机械研究所，慕尼黑技术大学）；University of Brasilia（巴西大学）；Carnegie Mellon University（卡内基梅隆大学）；Mechanical Engineering Program at New York University Abu Dhabi（纽约大学阿布扎克分校机械工程项目）；Mohamed bin Zayed University of Artificial Intelligence (MBZUAI)（穆罕默德·本·扎耶德人工智能大学）

Allen Emmanuel Binny, Mahathi Anand, Hugo T. M. Kussaba, Lingyun Chen, Shreenabh Agrawal, Fares J. Abu-Dakka, Abdalla Swikir

备注：: Accepted for publication in IEEE Robotics and Automation Letters (RA-L)
DOI：: 10.1109/LRA.2026.3655207

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文提出S$^2$-NNDS框架，通过神经网络学习复杂机器人运动，结合Lyapunov稳定性及屏障安全证书，提供概率保证，有效提升运动规划的安全性和稳定性。

摘要：

从演示中学习安全且稳定的机器人运动仍然是一个挑战，尤其是在涉及动态、障碍物丰富的复杂非线性任务中。在本文中，我们提出了Safe and Stable Neural Network Dynamical Systems S$^2$-NNDS，一种从演示中学习的框架，该框架同时学习具有表现力的神经动力学系统以及神经Lyapunov稳定性及屏障安全证书。与传统方法使用限制性的多项式参数化不同，S$^2$-NNDS利用神经网络来捕捉复杂的机器人运动，通过学习证书中的分割符合预测提供概率保证。在各种2D和3D数据集中的实验结果——包括LASA手写和从Franka Emika Panda机器人kinesthetically记录的演示——验证了S$^2$-NNDS在从可能不安全的演示中学习鲁棒、安全且稳定的运动方面的有效性。源代码、补充材料和实验视频可通过https://github.com/allemmbinn/S2NNDS访问。

英文摘要：

Learning safe and stable robot motions from demonstrations remains a challenge, especially in complex, nonlinear tasks involving dynamic, obstacle-rich environments. In this paper, we propose Safe and Stable Neural Network Dynamical Systems S$^2$-NNDS, a learning-from-demonstration framework that simultaneously learns expressive neural dynamical systems alongside neural Lyapunov stability and barrier safety certificates. Unlike traditional approaches with restrictive polynomial parameterizations, S$^2$-NNDS leverages neural networks to capture complex robot motions, providing probabilistic guarantees through split conformal prediction in learned certificates. Experimental results in various 2D and 3D datasets -- including LASA handwriting and demonstrations recorded kinesthetically from the Franka Emika Panda robot -- validate the effectiveness of S$^2$-NNDS in learning robust, safe, and stable motions from potentially unsafe demonstrations. The source code, supplementary material and experiment videos can be accessed via https://github.com/allemmbinn/S2NNDS

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。