Modeling the Locomotion of Articulated Soft Robots in Granular Medium

Yayun Du, Jacqueline Lam, Karunesh Sachanandani, Mohammad Khalid Jawed

TL;DR

This paper introduces a novel simulation framework for modeling the locomotion of articulated soft robots in granular media, combining discrete elastic rod simulations with resistive force theory. They built and tested a physical robot with rotating flagella, finding good agreement between simulations and experiments, and revealing counter-intuitive design principles like an optimal rotation speed for efficiency and a decrease in speed with more flagella.

What Problem Does This Paper Address?

Soft robots offer unique advantages in safety and adaptability, but their complex, deformable nature makes modeling their interaction with environments challenging. Specifically, understanding and predicting how these robots move through granular media (like sand or soil) is difficult. Unlike fluid dynamics, granular mechanics are less understood, and the interplay between a robot's elastic deformation and the granular medium's resistance is a significant hurdle for design and control. Existing models often struggle with the computational cost of simulating such complex interactions or don't fully capture the nuances of granular environments.

Methods

The core of this work is a custom numerical simulation tool. It represents the soft robot as a network of "Kirchhoff's rods," which are essentially flexible, elastic filaments. This discrete representation allows for detailed modeling of the robot's shape changes. The interaction with the granular medium is modeled using Resistive Force Theory (RFT), a common approach for low-Reynolds number fluid locomotion, adapted here for granular materials. RFT posits that the drag force on a filament is proportional to its velocity, with different coefficients for forces parallel and perpendicular to the filament.

The simulation uses a discrete differential geometry approach to track the robot's configuration (node positions and edge orientations) over time. It solves the equations of motion implicitly using a Newton-Raphson method, which is computationally efficient for this type of problem. A key innovation is how they handle the robot's actuation: the motor's rotation is incorporated by dynamically updating the "undeformed" state of the elastic rods, effectively simulating the applied strain.

Datasets

The paper does not explicitly mention pre-existing datasets for training or evaluation. Instead, they generated their own experimental data by building and testing a physical robot. This experimental data was then used to:

  1. Fit parameters for the simulation model (specifically, coefficients related to the head's drag and the granular medium's viscosity).
  2. Validate the simulation model by comparing its predictions against experimental results for different robot configurations (number of tails).

Key Insights and Ideas

  • RFT's surprising applicability (and limitations) in granular media: The authors found that RFT, typically used for fluids, provides a reasonable model for granular locomotion, especially with a small number of flagella. However, they observed that it breaks down when more flagella are added, suggesting the need for more advanced granular interaction models.
  • Counter-intuitive speed dependence on flagella count: Contrary to what one might expect, increasing the number of flagella from two to three decreased the robot's speed. This highlights the complex, nonlinear interplay between propulsion, drag, and the robot's internal mechanics.
  • Optimal rotational speed for efficiency: The research identified a specific rotational speed for the motor that maximizes the robot's efficiency (propulsive force per unit torque). This finding underscores the importance of precise control and design optimization for soft robots.
  • Rod-based modeling as a design tool: The simulation's speed (faster than real-time in some cases) makes it a valuable tool for exploring design trade-offs and optimizing robot performance before physical prototyping.

Key Results

  • Quantitative agreement: The simulation results showed good quantitative agreement with experimental data for robot speed and head rotation speed, particularly for robots with two and three flagella.
  • Parameter fitting: The simulation model was calibrated using experimental data from a two-tailed robot, yielding fitting parameters for head drag coefficients and granular viscosity.
  • Speed reduction with more flagella: For a fixed total rotational speed, the robot's speed decreased as the number of flagella increased from two to three.
  • Efficiency optimization: The study demonstrated a non-monotonic relationship between efficiency and total rotational speed, with an optimal speed identified around 150 rpm for a two-tailed robot.
  • Stick-slip dynamics observed: For robots with a larger number of flagella (four or five), experimental observations revealed "stick-slip" behavior, where the robot intermittently gets stuck. This phenomenon was not captured by the RFT model.

Limitations and Future Work

  • RFT limitations: The authors acknowledge that RFT fails to capture complex behaviors like "stick-slip" dynamics observed with a higher number of flagella. Future work could involve developing more sophisticated drag models for granular media that account for jamming and other nonlinear effects.
  • Head shape approximation: While they used empirical coefficients (C1, C2) to account for the non-spherical head shape in drag calculations, a more physically grounded model for head-granular interaction could improve accuracy.
  • Inertia neglected: The simulation assumes low Reynolds number dynamics, effectively neglecting inertia. While this simplifies the model and is justified by the dominance of viscous forces, it might limit applicability in scenarios with higher speeds or different media.
  • Future directions: The authors suggest improving the drag model for granular media and exploring how to integrate more complex physics into their simulation framework. They also highlight the potential of their fast simulation tool for parametric studies to optimize robot design and control strategies.

Summary updated

Abstract

We introduce a numerical tool for modeling articulated soft robots that couples discrete differential geometry-based simulation of elastic rods, our model for the articulated structure, and other external forces. Parallel to simulations, we build an untethered robot testbed, in the granular medium, comprised of multiple flexible flagella that are rotated about an axis by a motor. Drag from the granules causes the flagella to deform and the deformed shape generates a net forward propulsion. External drag depends on the flagellar shape, while the change in flagellar shape is the result of the competition between the external loading and elastic forces. We find reasonable quantitative agreement between experiments and simulations. Owing to a rod-based kinematic representation of the robot, the simulation can run faster than real-time in some cases, and, therefore, we can use it as a design tool for this class of soft robots. We find that there is an optimal rotational speed at which maximum efficiency is achieved. Moreover, both experiments and simulations show that increasing the number of flagella from two to three decreases the speed of the robot. This indicates that our simulator is potentially applicable for unknown physics exploration. We also gain insight into the mechanics of granular medium - while resistive force theory can successfully describe the propulsion at low number of flagella, it fails when more flagella are added to the robot.
BibTeX
@article{du2022modeling,
  author = {Yayun Du and Jacqueline Lam and Karunesh Sachanandani and Mohammad Khalid Jawed},
  title = {{Modeling the Locomotion of Articulated Soft Robots in Granular Medium}},
  journal = {IEEE Robotics Autom. Lett.},
  year = {2022},
  url = {https://openreview.net/forum?id=YTaEG1og2D},
}
Figure