SoPrA: Fabrication & Dynamical Modeling of a Scalable Soft Continuum Robotic Arm with Integrated Proprioceptive Sensing

Yasunori Toshimitsu, Ki Wan Wong, Thomas Buchner, Robert K. Katzschmann

TL;DR

This paper introduces SoPrA, a novel soft continuum robotic arm that integrates proprioceptive sensing with a physics-based dynamic model. This allows the robot to accurately predict its own movements and estimate external forces without relying on external cameras, paving the way for more capable and autonomous soft robots in complex environments.

What Problem Does This Paper Address?

Soft robots offer significant advantages in terms of compliance and safety, making them ideal for human interaction and delicate manipulation tasks. However, a major hurdle in their widespread adoption has been the difficulty in accurately modeling their complex, non-linear dynamics and integrating effective proprioceptive sensing. Existing research often focuses on individual components (actuators, sensors) rather than a holistic, integrated system. This paper addresses the need for a soft robotic arm that is not only scalable and manufacturable but also possesses an accurate analytical model and internal sensing capabilities to understand its own state and interaction with the environment, thereby reducing reliance on external perception systems.

Methods

The core of this work lies in the design and modeling of the SoPrA (Soft continuum Proprioceptive Arm).

  • Design: SoPrA is built using fiber-reinforced (FR) fluidic elastomer actuators, which provide controlled bending when pressurized. These actuators are arranged in segments, with each segment containing three FR chambers. Crucially, each segment also integrates a 2-axis digital flex sensor along its central axis. The arm is designed with geometric scalability in mind, allowing for variations in size and segment count.
  • Modeling: The paper proposes an analytical dynamic model based on the Piecewise Constant Curvature (PCC) approximation. To enhance accuracy, they increase the number of PCC elements per actuated segment (Npcc). The model accounts for inertial effects, material properties (stiffness and damping), and external forces. They employ an augmented rigid body model to derive kinetic parameters and an analytical model of the cross-section to calculate stiffness and damping.
  • Proprioception and Force Estimation: The integrated flex sensors provide proprioceptive measurements of the arm's pose. This data is then fused with the analytical dynamic model using a quadratic programming approach. This allows the system to estimate external contact forces without external sensors, by minimizing the discrepancy between the model's prediction and the sensor's measurement.

Datasets

Not discussed in the paper. The paper focuses on experimental validation using custom-built hardware and motion capture systems for ground truth.

Key Insights and Ideas

  • Integrated Proprioception is Key: The paper strongly advocates for integrating proprioceptive sensors directly into the soft robot's structure. This allows for a self-aware robot that can operate effectively even when external sensing is obstructed or unavailable.
  • PCC Model with Increased Npcc: While PCC is a standard approximation for soft robots, increasing the number of PCC elements per segment (Npcc) significantly improves model accuracy, offering a trade-off between computational cost and performance.
  • Model-Sensor Fusion for Force Estimation: Combining a physics-based dynamic model with proprioceptive sensor data provides a powerful mechanism for estimating external forces. This is a crucial step towards enabling soft robots to perform delicate manipulation tasks requiring force feedback.
  • Scalability and Manufacturability: The design emphasizes ease of fabrication and scalability, suggesting that SoPrA can be adapted for various applications and sizes, making it a practical platform for further research and development.

Key Results

  • Model Accuracy: The dynamic model, with Npcc=3, achieved an average tip position error of 1.27 cm for the upper segment and 1.96 cm for the lower segment during sinusoidal sweeps. Increasing Npcc to 3 improved accuracy by approximately 22.8% compared to Npcc=1.
  • External Load Estimation: The robot was able to estimate the relative weight of objects grasped by its gripper. While absolute weight estimation had some inaccuracies (e.g., estimating ~0.3N for 102g pliers), a linear relationship between actual and estimated weight was observed, indicating its utility for relative force sensing.
  • Robust Grasping: SoPrA could grasp various objects of different weights (17g to 117g) without requiring any physical modifications to the gripper.

Limitations and Future Work

  • Model Accuracy vs. Npcc: While increasing Npcc improves accuracy, it also increases computational cost. The paper notes that further increases beyond Npcc=3 did not yield significant improvements and even led to a slight increase in error.
  • Absolute Force Estimation: The paper acknowledges that the absolute estimation of external forces was not perfectly accurate, potentially due to the material property characterization being done under actuated conditions rather than known external loads.
  • Dynamic Actuation Modeling: For very fast movements, a more detailed model of the pneumatic supply might be necessary.
  • Future Work: The authors plan to continue improving the design and model formulation, explore more modes of proprioception, and investigate the application of SoPrA to practical tasks in everyday environments. They also suggest exploring adaptive models that can dynamically adjust Npcc to minimize computational complexity.

Summary updated

Abstract

Due to their inherent compliance, soft robots are more versatile than rigid linked robots when they interact with their environment, such as object manipulation or biomimetic motion, and are considered to be the key element in introducing robots to everyday environments. Although various soft robotic actuators exist, past research has focused primarily on designing and analyzing single components. Limited effort has been made to combine each component to create an overall capable, integrated soft robot. Ideally, the behavior of such a robot can be accurately modeled, and its motion within an environment uses its proprioception, without requiring external sensors. This work presents a design and modeling process for a Soft continuum Proprioceptive Arm (SoPrA) actuated by pneumatics. The integrated design is suitable for an analytical model due to its internal capacitive flex sensor for proprioceptive measurements and its fiber-reinforced fluidic elastomer actuators. The proposed analytical dynamical model accounts for the inertial effects of the actuator’s mass and the material properties, and predicts in real-time the soft robot’s behavior. Our estimation method integrates the analytical model with proprioceptive sensors to calculate external forces, all without relying on an external motion capture system. SoPrA is validated in a series of experiments demonstrating the model’s and sensor’s accuracy in estimation. SoPrA will enable soft arm manipulation including force sensing while operating in obstructed environments that disallows exteroceptive measurements.
BibTeX
@inproceedings{toshimitsu2021sopra,
  author = {Yasunori Toshimitsu and Ki Wan Wong and Thomas Buchner and Robert K. Katzschmann},
  title = {{SoPrA: Fabrication \& Dynamical Modeling of a Scalable Soft Continuum Robotic Arm with Integrated Proprioceptive Sensing}},
  booktitle = {IEEE/RSJ International Conference on Intelligent Robots and Systems},
  year = {2021},
  url = {https://openreview.net/forum?id=3PeCxYjxrx},
}
Figure