# Φ-CON模型实现双臂机器人柔性线状物动力学学习与稳定控制

> 原文链接：https://robotics.hellobike.com/papers/2026-11-19-2611-23563-e84586a0
> 频道：papers | 更新日期：2026-11-19 | 标签：auto, arxiv
> 核心摘要：针对刚性重物操作难题，研究提出Φ-CON耦合振荡器网络学习力驱动力学。仿真参数量较NODE减少97.2%，真机双臂实验端点调节误差降低27.9%，通过拉格朗日结构保障闭环稳定性。

# Φ-CON模型实现双臂机器人柔性线状物动力学学习与稳定控制

::: v-pre
> 本文由 AI 根据公开内容整理，可能有误，请以原文为准：[Learning the Dynamics of Deformable Linear Objects Through Coupled Oscillators for Bimanual Robotic Manipulation](https://arxiv.org/abs/2611.23563)（arXiv 2611.23563）

## 核心内容

针对现有柔性线状物（DLO）学习方法忽略动态效应与机械响应、难以处理刚重物体的局限，研究提出Φ-CON模型以学习力驱动动力学。该方法在四种仿真DLO及真实双臂机器人平台上验证，相比无模型PD基线，仿真中端点调节误差降低97.5%至99.7%，真机双端点平均误差减少27.9%，证实了显式建模机械响应对刚重物体操控的关键作用。

技术上，Φ-CON结合浮基耦合振荡器网络（FBCON）与可仿射微分同胚映射$\Phi$，其形状动力学由构造保证稳定，拉格朗日结构提供动能与势能显式表达。基于此设计的模型PD控制器补偿学习到的势力，确保闭环收敛。相比神经ODE与拉格朗日神经网络，该模型参数量分别减少97.2%和90.2%，且预测精度相当，显著提升了计算效率与物理一致性。

## 关键看点

- **参数效率**：Φ-CON预测精度对标NODE与LNN，但参数量分别减少97.2%和90.2%，大幅降低模型复杂度
- **控制性能**：基于拉格朗日结构的模型PD控制器，使仿真端点调节误差较无模型基线降低97.5%-99.7%
- **真机验证**：在真实双臂机器人实验中，双端点平均调节误差减少27.9%，验证了对刚重DLO的操控能力
- **理论保障**：通过浮基耦合振荡器网络与仿射微分同胚映射，从构造上保证形状动力学稳定与闭环收敛

## 工程与产业观察

该工作为具身智能操作非结构化柔性物体提供了物理一致的动力学建模范式。通过将拉格朗日力学归纳偏置嵌入神经网络，在保证小样本学习效率的同时获得闭环稳定性证明，解决了传统黑盒模型在刚重线缆装配等精密任务中泛化差、安全性低的工程痛点，对工业级双臂操作算法研发具有直接参考价值。

## 英文摘要 (Abstract)

Many learning-based approaches for deformable linear objects (DLOs) rely on simplified deformation models that neglect dynamic effects and the mechanical response of the objects, limiting their applicability to lightweight, compliant objects and slow motions. Inspired by recent advances in learning physically structured dynamics with coupled oscillator networks (CONs), we propose $\Phi$-CON for learning force-driven DLO dynamics. The model combines a floating-base coupled oscillator network (FBCON), whose latent potential energy depends only on the DLO's shape, with a learned affine diffeomorphism $\Phi$ between observed positions along the DLO and latent coordinates. Its shape dynamics are stable by construction, and its Lagrangian structure yields explicit expressions for kinetic and potential energy. We exploit this structure to design a model-based proportional-derivative (PD) controller that compensates for the learned potential forces, thereby establishing closed-loop stability guarantees and ensuring convergence. We evaluate the approach on four simulated DLOs and a real bimanual robot setup. In simulation, $\Phi$-CON achieves prediction accuracy comparable to neural ordinary differential equation (NODE) models and Lagrangian neural networks (LNNs), with $97.2\%$ and $90.2\%$ fewer parameters, respectively. Relative to a model-free PD baseline, the proposed controller reduces mean endpoint regulation error by $97.5$-$99.7\%$ in simulation and achieves a mean per-target reduction of $27.9\%$ across both endpoints in real-world experiments, demonstrating the importance of explicitly accounting for the mechanical response of stiff and heavy objects.

作者：Iv\'an L\'opez-Broce\~no, Maximilian St\"olzle, Daniel Feliu-Talegon, Daniela Rus, Cosimo Della Santina
:::

## 原始参考来源 (Sources)

- [arXiv 2611.23563：Learning the Dynamics of Deformable Linear Objects Through Coupled Oscillators for Bimanual Robotic Manipulation](https://arxiv.org/abs/2611.23563)