Certifiably Safe Manipulation of Deformable Linear Objects via Joint Shape and Tension Prediction
Yiting Zhang, Shichen Li · 2025
Read Certifiably Safe Manipulation of Deformable Linear Objects via Joint Shape and Tension Prediction by Yiting Zhang, Shichen Li (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-13889.txt.
Introduction
Robots that handle cables and wires must predict how the object will bend and how hard it will pull, or they risk damage. Yiting Zhang and Shichen Li present a planning framework in this 2025 paper that forecasts both shape and tension together.
Those forecasts feed a real-time optimizer built on set-based math that can certify safety along the whole path. Tested on a simulated wire harness assembly with a seven-joint arm, the system finished more tasks than leading rivals and never broke a safety rule. The authors say joint shape-and-tension prediction is the key to gentle, reliable handling of floppy objects.
Catalog record
| CREATOR | Yiting Zhang, Shichen Li |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 24 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-13889
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Certifiably Safe Manipulation of Deformable Linear Objects via Joint Shape and Tension Prediction Yiting Zhang, Shichen Li Abstract Manipulating deformable linear objects (DLOs) is challenging due to their complex dynamics and the need for safe interaction in contact-rich environments. Most existing models focus on shape prediction alone and fail to account for contact and tension constraints, which can lead to damage to both the DLO and the robot. In this work, we propose a certifiably safe motion planning and control framework for DLO manipulation. At the core of our method is a predictive model that jointly estimates the DLO’s future shape and tension. These predictions are integrated into a real-time trajectory optimizer based on polynomial zonotopes, allowing us to enforce safety constraints throughout the execution. We evaluate our framework on a simulated wire harness assembly task using a 7-DOF robotic arm. Compared to state-of-the-art methods, our approach achieves a higher task success rate while avoiding all safety violations. The results demonstrate that our method enables robust and safe DLO manipulation in contact-rich environments. I Introduction Effective manipulation of deformable linear objects (DLOs)—such as cables, ropes, and wires—is essential for a wide range of applications, including wire harness assembly and surgical suturing [ 1 , 2 , 3 ] . Yet, DLO manipulation remains a fundamental challenge in robotics due to their high-dimensional configuration spaces and nonlinear, time-varying dynamics [ 4 ] . Moreover, in real-world applications, contact between the DLO and its environment is often inevitable and sometimes even necessary [ 5 ] . Such interactions introduce additional safety risks, including overextension and potential damage to both the DLO and the robot [ 6 ] . These challenges underscore the need for a manipulation framework that integrates high-fidelity modeling with planning and control strategies equipped with safety guarantees. Existing modeling approaches for DLOs primarily focus on predicting shape deformations and can be broadly categorized into physics-based and learning-based methods. Physics-based models, such as mass-spring systems [ 7 ] , position-based dynamics [ 8 ] , and the finite element method [ 9 ] , offer physical interpretability but often suffer from a trade-off between high computational cost and limited accuracy in dynamic scenarios. Learning-based methods leverage deep neural networks to predict DLO shapes and sometimes incorporate physics priors to enhance consistency [ 10 , 11 , 12 , 13 ] . While these models have achieved promising results in shape prediction, they face two key limitations: they generalize poorly to environments with frequent contact, and they cannot assess whether the DLO will be overextended—an essential factor in ensuring safe manipulation. Similar challenges have also been observed in structured manipulation settings involving cluttered scenes [ 14 ] . Furthermore, shape prediction alone is insufficient; accurate tension estimation is critical for preventing failure and enabling reliable execution in safety-critical tasks. In real-world manipulation scenarios, however, tension information is rarely directly observable and is often difficult to model due to frictional contact, geometric constraints, and limited sensing [ 15 ] . These challenges make it difficult to analytically infer internal tension from geometric information alone. However, learning-based methods have been effectively applied in related contexts to approximate unobservable physical quantities [ 16 , 17 , 18 ] , suggesting their potential to address this gap. Fig. 1: This paper introduces a certifiably safe framework for manipulating DLOs in contact-rich environments. At the core of our method is a predictive model that simultaneously estimates the future shape and tension of the DLO. The figure illustrates a robot arm manipulating a DLO (white) toward a goal configuration (green) in the presence of an obstacle (red). The manipulation is executed without causing collisions or overstretching, despite contact with the environment. On the planning side, existing methods often rely on simplifying assumptions that compromise safety. Sampling-based approaches, such as probabilistic roadmaps (PRM) [ 19 ] and rapidly-exploring random trees (RRT) [ 20 , 21 ] , as well as learning-based methods using supervised [ 22 ] or reinforcement learning [ 23 ] , typically neglect contact-induced risks by assuming no overextension and no collisions between the robot and its environment. However, in real-world scenarios, a DLO can become trapped or overstretched, generating unexpected forces that alter the robot’s behavior and potentially lead to unanticipated collisions. These safety hazards underscore the need for a motion planning and control framework that not only accounts for DLO deformation but also certifiably enforces safety under physical constraints. Fig. 2: Overview of the proposed framework. Starting from an initial state, the robot generates a set of parameterized trajectories and uses a learned predictive model to estimate the resulting DLO shape and tension. These predictions are incorporated into a reachable set computation, enabling the identification of safe trajectories under contact-rich conditions. An optimization problem is then solved to select the best feasible trajectory, which is executed using a robust controller. The loop runs in a receding horizon fashion until the DLO reaches the desired goal configuration without collisions or overstretching. In this work, we propose a novel learning-based predictive model that jointly forecasts the future shape and tension of DLOs over extended time horizons. Building on this model, we develop a certifiably safe motion planning and control framework for DLO manipulation in complex, 3D contact-rich environments. The framework explicitly enforces safety constraints on DLO tension and robot motion through an online trajectory optimization process. This enables robust motion execution without inducing excessive tension or collisions. Experimental results in a simulated wire harness assembly task demonstrate substantial improvements in accuracy, safety, and robustness compared to existing baselines. II Methods This paper addresses the problem of safe motion planning and control for DLO manipulation in contact-rich environments. We consider scenarios in which one end of the DLO is held by the robot’s end-effector, while the other end is either fixed to the environment or held stationary by another robot. The objective is to develop a framework that leverages predictions of DLO shape and tension to guide the robot in manipulating the DLO from an initial configuration to a desired one. The proposed approach aims to prevent both manipulator collisions and DLO overextension. An overview of the framework is shown in Fig. 2 . II-A DLO Shape and Tension Prediction In this subsection, we present our approach to jointly predicting the shape and tension of a DLO during dynamic manipulation. We begin by introducing a discretized representation of the DLO. We then leverage the structure of a long short-term memory (LSTM) network to develop a model that predicts future DLO shape and tension. The model requires only the current DLO state and the planned future states of the manipulator as inputs. II-A 1 Discretized state representation Fig. 3 illustrates the discretized state representation for the DLO and shows the local coordinates for each discretized node. We define the state of the DLO at time t as \mathbf{x}(t) . We discretize the DLO into N small links with N+1 nodes so the state of entire DLO can be represented by the state of each node: \textbf{x}(t)=[x_{0}(t),x_{1}(t),...,x_{N}(t)]^{\top} . Each node’s state can be explicitly denoted by its position in the world coordinates: x_{i}(t)\in\mathbb{R}^{3} for i=0,1,...,N . Therefore, the DLO’s state can be denoted by: x\in\mathbb{R}^{N+1} . Neighbor DLO