Model-Based Disturbance Estimation for a Fiber-Reinforced Soft Manipulator using Orientation Sensing

Barnabas Gavin Cangan, Stefan Escaida Navarro, Bai Yang, Yu Zhang, Christian Duriez, Robert K. Katzschmann

Abstract

For soft robots to operate effectively in human-centered environments, they must possess the ability to estimate their state and external interactions using proprioceptive sensors. This paper presents a model-based approach for estimating external disturbances on a multi-segment soft continuum robotic arm. By leveraging the Finite Element Method (FEM) within the SOFA framework, we developed a high-fidelity digital twin that accounts for complex material deformations and fiber-reinforced pneumatic actuation. Using sparse orientation data, we formulate a quadratic programming (QP) inverse dynamics problem to calculate external forces. Experimental results demonstrate a remarkably low average force estimation error of approximately 1.2%, suggesting that high-fidelity modeling can enable reactive, sensor-based behavior in compliant robotic systems.

Introduction

Soft robots offer inherent safety advantages for human-robot interaction due to their compliant structures. However, this same flexibility introduces significant challenges for state estimation and control. To perform precise tasks or interact safely with their surroundings, these robots must detect external forces—a process known as disturbance estimation.

Traditional approaches often rely on simplified kinematics, such as Piece-wise Constant Curvature (PCC) approximations or matched rigid-body models. While computationally efficient, these models frequently fail to capture the non-linear deformations and anisotropic behaviors of soft materials, particularly when reinforced with fibers. While data-driven methods offer an alternative, they often lack the generalization required to handle environmental changes or novel configurations. This research addresses these gaps by employing a high-fidelity FEM-based model to achieve accurate force estimation without the need for dense sensor arrays.

Methodology

High-Fidelity FEM Modeling

The researchers developed a digital twin of the "SoPrA" arm—a multi-segment soft manipulator featuring six internal pneumatic chambers. A critical aspect of this model is the representation of fiber reinforcement. In physical soft robots, fibers are used to prevent pneumatic chambers from bulging outward, directing expansion into longitudinal motion. The authors modeled these fibers as a series of stiff springs mapped directly to the FEM mesh within the SOFA (Simulation Open Framework Architecture) environment, providing a computationally efficient yet accurate representation of the arm's structural constraints.

State Observation and Inverse Dynamics

Rather than tracking the entire surface of the robot, the system utilizes a state observer based on sparse orientation data. By measuring the orientation of rigid sections at the end of each segment (using markers or sensors), the model can infer the overall pose.

To estimate external forces, the team formulated an inverse dynamics problem using Quadratic Programming (QP). The solver identifies the external force components that minimize the discrepancy between the robot's predicted pose (based on internal pressures) and its actual sensed orientation. This allows the system to "attribute" pose errors to external disturbances.

Calibration

Recognizing that manual fabrication processes introduce imperfections, a two-stage calibration process was implemented. This involved tuning the material properties (specifically Young’s Modulus) and pressure scaling factors to ensure the FEM model accurately reflected the physical hardware's behavior under various loads.

Experimental Results

The proposed method was validated using a physical SoPrA arm monitored by a Qualisys motion capture system for ground-truth orientation and a 1-axis force gauge for force validation.

  • Force Estimation Accuracy: The system achieved an average force estimation error of 1.2%. In unpressurized scenarios, the mean force error was a negligible 0.007 N, while in more complex pressurized states, the error remained low at 0.0238 N.
  • Teach-In Mode: The researchers demonstrated a "teach-in" application where a human manually guides the robot. The simulation calculates the pressures required to maintain the user-defined pose. In tests, this mode showed a mean position error of only 5.2 mm and an orientation error of 4.04°.
  • Computational Performance: The FEM simulation achieved a refresh rate of approximately 12 Hz. While not yet at high-speed control rates, this performance is sufficient for many human-robot interaction tasks and approaches the threshold for real-time closed-loop applications.

Discussion and Conclusion

This research demonstrates that sparse orientation sensing, when combined with a high-fidelity FEM model, is sufficient for accurate disturbance estimation in soft manipulators. The ability to model fiber reinforcement as discrete elements allows for a balance between physical accuracy and computational speed.

Limitations and Future Work

Despite the high accuracy, the current system relies on external motion capture for orientation data. Future iterations aim to integrate embedded Inertial Measurement Units (IMUs) to achieve true proprioception. Additionally, the current model assumes forces are applied at the segment tips; developing the ability to localize contact points along the entire length of the arm—perhaps through the addition of a tactile "sensor skin"—remains a key objective for future research.

Ultimately, these results provide a robust foundation for building reactive soft robots capable of operating safely and intelligently in complex, human-centered environments.

Summary updated

Abstract

For soft robots to work effectively in human-centered environments, they need to be able to estimate their state and external interactions based on (proprioceptive) sensors. Estimating disturbances allows a soft robot to perform desirable force control. Even in the case of rigid manipulators, force estimation at the end-effector is seen as a non-trivial problem. And indeed, other current approaches to address this challenge have shortcomings that prevent their general application. They are often based on simplified soft dynamic models, such as the ones relying on a piece-wise constant curvature (PCC) approximation or matched rigid-body models that do not represent enough details of the problem. Thus, the applications needed for complex human-robot interaction can not be built. Finite element methods (FEM) allow for predictions of soft robot dynamics in a more generic fashion. Here, using the soft robot modeling capabilities of the framework SOFA, we build a detailed FEM model of a multi-segment soft continuum robotic arm composed of compliant deformable materials and fiber-reinforced pressurized actuation chambers with a model for sensors that provide orientation output. This model is used to establish a state observer for the manipulator. Model parameters were calibrated to match imperfections of the manual fabrication process using physical experiments. We then solve a quadratic programming inverse dynamics problem to compute the components of external force that explain the pose error. Our experiments show an average force estimation error of around 1.2%. As the methods proposed are generic, these results are encouraging for the task of building soft robots exhibiting complex, reactive, sensor-based behavior that can be deployed in human-centered environments.
BibTeX
@inproceedings{cangan2022model,
  author = {Barnabas Gavin Cangan and Stefan Escaida Navarro and Bai Yang and Yu Zhang and Christian Duriez and Robert K. Katzschmann},
  title = {{Model-Based Disturbance Estimation for a Fiber-Reinforced Soft Manipulator using Orientation Sensing}},
  booktitle = {IEEE/RSJ International Conference on Intelligent Robots and Systems},
  year = {2022},
  url = {https://openreview.net/forum?id=cBsO04Evzx},
}
Figure