RopeDreamer: A Kinematic Recurrent State Space Model for Dynamics of Flexible Deformable Linear Objects

2026-04-30

AI Summary

This work addresses the challenge of modeling deformable linear objects (DLOs) in robotic manipulation, where high-dimensional nonlinear dynamics and dense contacts often lead to self-intersections and non-physical deformations. To this end, the authors propose a latent dynamics framework based on a recurrent state-space model, incorporating a novel quaternion-based kinematic chain representation. By modeling relative rotations instead of Cartesian coordinates, the approach inherently preserves constant link lengths and constrains the configuration to a physically valid manifold. A dual-decoder architecture is employed to decouple state reconstruction from future prediction, encouraging the latent space to better capture true dynamics. Experiments demonstrate that, over 50-step open-loop predictions, the method reduces prediction error by 40.52% and inference time by 31.17% compared to the state-of-the-art, while significantly improving topological consistency and physical plausibility in multi-crossing scenarios.

Abstract

The robotic manipulation of Deformable Linear Objects (DLOs) is a fundamental challenge due to the high-dimensional, non-linear dynamics of flexible structures and the complexity of maintaining topological integrity during contact-rich tasks. While recent data-driven methods have utilized Recurrent and Graph Neural Networks for dynamics modeling, they often struggle with self-intersections and non-physical deformations, such as tangling and link stretching. In this paper, we propose a latent dynamics framework that combines a Recurrent State Space Model with a Quaternionic Kinematic Chain representation to enable robust, long-term forecasting of DLO states. By encoding the DLO as a sequence of relative rotations (quaternions) rather than independent Cartesian positions, we inherently constrain the model to a physically valid manifold that preserves link-length constancy. Furthermore, we introduce a dual-decoder architecture that decouples state reconstruction from future-state prediction, forcing the latent space to capture the underlying physics of deformation. We evaluate our approach on a large-scale simulated dataset of complex pick-and-place trajectories involving self-intersections. Our results demonstrate that the proposed model achieves a 40.52% reduction in open-loop prediction error over 50-step horizons compared to the state-of-the-art baseline, while reducing inference time by 31.17%. Our model further maintains superior topological consistency in scenarios with multiple crossings, proving its efficacy as a compositional primitive for long-horizon manipulation planning.

Problem

• Deformable Linear Objects
• non-physical deformations
• self-intersections
• topological integrity
• dynamics modeling

Innovation

• Recurrent State Space Model
• Quaternionic Kinematic Chain
• Deformable Linear Objects
• Topological Consistency
• Latent Dynamics Modeling