Skip to content

Contains data and code for the paper Untethered Soft Robots - Design and Closed Loop Motion Planning with Discrete Elastic Rods

Notifications You must be signed in to change notification settings

softmachineslab/frog

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 
 
 

Repository files navigation

[Untethered Soft Robots - Design and Closed Loop Motion Planning with Discrete Elastic Rods]

Numerical simulation for untethered soft robots using DER method.



How to Use

Dependencies

Install the following C++ dependencies:

  • Eigen

    • Eigen is used for various linear algebra operations.
  • Pardiso

    • Pardiso is used for the solving of linear systems.
  • OpenGL / GLUT

    • OpenGL / GLUT is used for rendering the knot through a simple graphic.
    • Simply install through apt package manager:
      sudo apt-get install libglu1-mesa-dev freeglut3-dev mesa-common-dev
  • Lapack (usually preinstalled on your computer)


Compiling

g++ -I 'Eigen link' main.cpp world.cpp elasticRod.cpp setInput.cpp timeStepper.cpp inertialForce.cpp externalGravityForce.cpp elasticStretchingForce.cpp geometry.cpp elasticBendingForce.cpp dampingForce.cpp -lGL -lglut -lGLU -lpardiso600-GNU720-X86-64 -llapack -lgfortran -fopenmp -lpthread -lm -Ofast -o simDER



Geometry

The input file for geometry can be generated by "geometry.py" or "geometry.cpp".

  • nodes.txt - Position of input nodes, N*2, [node x position, node y position].
  • stretching.txt - Stretching element, N*2, [nodeIndex1, nodeIndex2].
  • bending.txt - Bending element, N*4, [nodeIndex1, nodeIndex2, nodeIndex3, relative stiffness].
  • actuationSequence.txt - Actuation sequence of four limbs. "1" represents activation and "0" represents no activation
  • actuation.txt - Nodes of four limbs. Each column corresponds to a single limb.

Setting Parameters

All simulation parameters are set through a parameter file option.txt.

Specifiable parameters are as follows (we use SI units):

  • RodLength - Contour length of the rod.
  • rodCurvature - Undeformed curvature of rod.
  • rodDensity - Density of rod on a limb.
  • frameDensity - Density of rod on the frame.
  • deltaLength - Length of each rod.
  • rodRadius - Radius of the rod.
  • frameWidth - Width of the frame.
  • frameLength - Length of the frame.
  • limbNumber - Number of limbs.
  • youngM - Young's modulus of rod.
  • Poisson - Poisson ratio of rod.
  • deltaTime - Time of each step in the simulation.
  • totalTime - Total time in the simulation.
  • tol - Tolerance for convergence.
  • maxIteration - Max iteration for convergence.
  • gVector - Gravity vector.
  • render - Render option. "1" represents "on" and "0" represents "off".
  • saveData - Save data option. "1" represents "on" and "0" represents "off".
  • cf1 - Drag coefficient of the frame in orthogonal direction.
  • cf2 - Drag coefficient of the frame in tangent direction.
  • cr1 - Drag coefficient of the limb in orthogonal direction.
  • cr2 - Drag coefficient of the limb in tangent direction.
  • phaseDelay - Phase offset between front limbs and rear limbs.
  • frequency - Actuation frequency of limbs.
  • tauLogisitic - Coefficient in the equation characterizing Young's modulus change of limbs during actuation and deactivation.
  • actuationTime - Actuation time of each limb.
  • actA - actD - Coefficient in the equation characterizing curvature change of limbs during actuation.
  • deactA - deactD - Coefficient in the equation characterizing curvature change of limbs during deactuation.
  • addMass - Coefficient of added mass.
  • EGaIn - EGaIn volume ratio in "%".

Fitting data for actuation

  • fitF00_act - fitF50_act - Coefficient in the equation characterizing curvature change of the limb during actuation at each frequency for a corresponding volume ratio of EGaIn.
  • fitF00_deact - fitF50_deact - Coefficient in the equation characterizing curvature change of the limb during deactuation at each frequency for a corresponding volume ratio of EGaIn.


Running the Simulation

Once parameters are set to your liking, the simulation can be run from the terminal by running the provided script:

./simDER option.txt



Saving the Data

Make sure you set the option to save the data from your simulations. You can easily run it in a loop in a python script:

#!/usr/bin/python3

import os

num_tests = 2000

for i in range(num_tests):
    print('Starting test: ' + str(i))
    os.system("./simDER option.txt")

Once this data is collected, it can be processed into a .npy file for direct implementation in numpy.



Planning/Control Software


Dependencies

To setup the environment, make the 'conda_setup.bash' file executable, then run it. Say yes to its prompts, then follow the text instructions it spits out.

To make conda_setup executable, run

$chmod+x conda_setup.bash

$./conda_setup.bash

in the folder that conda_setup.bash is found in.


Running the Planner

Once a data library is collected, the controller can be run from the terminal by running:

./frog_planner.

This script simply calls:

python3 src/main.py 'False' 'brute' 'si' 1 'data/simfrog_numpy_data_2022-01-26.npy' '/data/results'.

Options can be adjusted as follows:

  • Option 1: T/F if there is a calibration grid. The code works well without it, so I use 'False'.
  • Option 2: Planner uses nearest neighbor ('brute') or locally weighted regression ('lwr') to determine the next state.
  • Option 3: Type of path to follow. Options are line 'li', sinusoid 'si', and ellipsoid 'el'.
  • Option 4: Depth of tree of the planner (integer). Works well with smaller values (e.g. 2).
  • Option 5: is the numpy file containing the transition models from the DER library (e.g. 'data/simfrog_numpy_data_2022-01-26.npy').
  • Option 6: directory of saved results.

About

Contains data and code for the paper Untethered Soft Robots - Design and Closed Loop Motion Planning with Discrete Elastic Rods

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published