Unified Kinematic and Dynamical Modeling of a Soft Robotic Arm by a Piecewise Universal Joint Model

Zhanchi Wang, Gaotian Wang, Nikolaos M. Freris

TL;DR

This paper introduces a new "Piecewise Universal Joint" (PUJ) model for soft robotic arms that overcomes the limitations of the traditional Piecewise Constant Curvature (PCC) model, especially when dealing with external forces and gravity. The PUJ model offers significantly improved accuracy in both kinematic and dynamic predictions for 3D soft arm movements.

What Problem Does This Paper Address?

Soft robotic arms are incredibly versatile, enabling applications from human-robot interaction to surgery. However, their inherent flexibility and continuous deformation make them notoriously difficult to model accurately. The most common approach, Piecewise Constant Curvature (PCC), assumes each segment of the arm bends into a simple arc. While useful, this assumption breaks down when external forces, like gravity or interaction forces, are present, as the arm's deformation becomes more complex and non-constant. This leads to significant errors in predicting the arm's position and motion, hindering precise control. This paper aims to bridge this gap by developing a more robust and accurate modeling framework for 3D soft robotic arms that can handle these challenging scenarios.

Methods

The core innovation is the Piecewise Universal Joint (PUJ) model. Instead of assuming constant curvature, each segment of the soft arm is modeled as a stretchable link connected by a universal joint. This allows for more complex planar deformations within each segment without twisting.

Here's a breakdown of their approach:

  • Kinematic Modeling: The soft arm is abstracted as a series of universal joints, similar to how rigid serial robots are modeled. They use homogeneous transformation matrices to represent the relationship between coordinate frames at the ends of each segment. The configuration variables for each universal joint are defined as translation lengths and two rotation angles. They also introduce an "Exponential Coordinate Representation" using the Product of Exponentials (PoE) formula, a standard technique in rigid robot kinematics, to unify the kinematic and dynamic descriptions.
  • Dynamic Modeling: They extend rigid-body dynamics principles to the PUJ model. The dynamics of each universal joint segment are derived using a recursive Newton-Euler method. Crucially, they account for the viscoelasticity of soft materials by adding linear stiffness (K) and damping (D) terms to the standard rigid-body dynamic equation.
  • System Identification: To make the model practical, they detail how to estimate the model parameters (M, C, G, K, D) from experimental data. This involves measuring the arm's physical properties and then using system identification techniques on free-swinging motion data to determine stiffness and damping coefficients.
  • Torque-to-Pressure Mapping: For their specific pneumatic soft arm, they establish a conversion from air pressure inputs to the equivalent torques used in the dynamic model.

Datasets

The paper uses a custom-built, four-segment pneumatic soft robotic arm for experimental validation. The "dataset" consists of motion capture data collected from this real robot platform under various conditions (different pressures, with and without load). This data is used to:

  • Validate the kinematic model by comparing predicted end-effector poses against measured poses.
  • Validate the dynamic model by comparing predicted joint trajectories against measured joint trajectories during free swing and step response experiments.

No new public datasets were introduced.

Key Insights and Ideas

  • Universal Joints for Softness: The key insight is that modeling soft arm segments as universal joints, rather than simple arcs, provides a much richer representation of their deformation capabilities, especially under external forces. This is a significant departure from the widely adopted PCC model.
  • Bridging Rigid and Soft Robotics: By adapting established rigid-body kinematic and dynamic formalisms (like PoE and Newton-Euler) to their PUJ model, the authors effectively leverage decades of research in rigid robotics for soft robots. This makes their approach more accessible and potentially easier to extend.
  • PUJ's Superiority in Real-World Scenarios: The experiments clearly demonstrate that the PUJ model significantly outperforms the PCC model when gravity or external loads are present. This is critical for practical applications where such forces are unavoidable.
  • Viscoelasticity Matters: Explicitly incorporating stiffness and damping terms into the dynamic model is essential for accurately capturing the behavior of soft materials, which are inherently viscoelastic.

Key Results

  • Kinematic Accuracy: The PUJ model achieved significantly lower maximum position and rotation errors compared to the PCC model, with errors being 6.7x lower for position and 4.6x lower for rotation when considering gravity and external forces.
  • Dynamic Accuracy (Free Swing): In free-swinging experiments (with and without a 200g load), the PUJ model showed high accuracy, with average joint angle errors of 0.57° (no load) and 1.1° (with load).
  • Dynamic Accuracy (Step Response): For step response experiments, the PUJ model demonstrated good accuracy, with average joint errors ranging from 1.6° to 2.5° depending on the scenario (on-plane, out-of-plane, with load).
  • Model Independence from Pressure: A notable finding is that the error of the PUJ model does not significantly increase with pressure (i.e., deformation), making it robust for scenarios with external forces.

Limitations and Future Work

  • Inextensibility Assumption: The authors acknowledge that their current PUJ model assumes segments are inextensible. They note that this can lead to some discrepancies, particularly in predicting more oscillatory behavior than observed in reality, as pressure can cause actual extension.
  • Planar Deformation: The model assumes each segment deforms in a plane without twisting. While this simplifies the model, real soft arms might exhibit some degree of twisting.
  • Future Work: The authors plan to extend their work to develop dynamical models for extensible arms based on the universal joint principle. They also aim to leverage their developed model for controlling soft robotic arms.

Summary updated

Abstract

The compliance of soft robotic arms renders the development of accurate kinematic & dynamical models especially challenging. The most widely used model in soft robotic kinematics assumes Piecewise Constant Curvature (PCC). However, PCC fails to effectively handle external forces, or even the influence of gravity, since the robot does not deform with a constant curvature under these conditions. In this paper, we establish three-dimensional (3D) modeling of a multi-segment soft robotic arm under the less restrictive assumption that each segment of the arm is deformed on a plane without twisting. We devise a kinematic and dynamical model for the soft arm by deriving equivalence to a serial universal joint robot. Numerous experiments on the real robot platform along with simulations attest to the modeling accuracy of our approach in 3D motion with load. The maximum position/rotation error of the proposed model is verified 6.7x/4.6x lower than the PCC model considering gravity and external forces.
BibTeX
@misc{wang2021unified,
  author = {Zhanchi Wang and Gaotian Wang and Nikolaos M. Freris},
  title = {{Unified Kinematic and Dynamical Modeling of a Soft Robotic Arm by a Piecewise Universal Joint Model}},
  year = {2021},
  eprint = {2109.05791},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2109.05791},
}
Figure