TacIPC: Intersection- and Inversion-Free FEM-Based Elastomer Simulation for Optical Tactile Sensors

Requirements

• CMake >= 3.20
• gcc/g++ >= 10
• SuiteSparse = 7.3.1
• xrandr
• xinerama
• xcursor
• xi
• OpenGL
• zlib

# install with apt
sudo apt install libsuitesparse-dev
sudo apt install libxrandr-dev
sudo apt install libxinerama-dev 
sudo apt install libxcursor-dev
sudo apt install libxi-dev
sudo apt install libgl1-mesa-dev
sudo apt install zlib1g-dev

Configure && Build

Build the project, and the executable will be generated in the build directory.

cmake -Bbuild build # configure
cmake --build build --parallel [N_threads] # build

Run example experiments

Run with Python script (headless by default)

1. Run the experiment

   1. Run experiment script

      python scripts/run_experiments.py

   2. Check out the experiment result directory output/experiments/[timestamp], and the log files in output/logs/[expName].log.
      • output/experiments/[timestamp]/config.json contains the experiment configuration.
      • output/experiments/[timestamp]/objs/ contains
        • the deformed object meshes ([objName]_[frameId].obj)
        • the object state files ([objName]_[frameId].json)

2. Postprocess and obtain marker displacements

   1. Create output/experiments_markers directory if not exists.
   2. Move/Copy the result directories into output/experiments_markers.
   3. Run the postprocess script

      python process/gen_marker_disp.py

   4. Check out the marker displacement files in output/experiments_markers/[expName]/markers/.

3. (optional) Visualize the marker displacements

   1. Run the visualization script

      python process/vis_marker_disp.py

   2. Check out the visualization in output/experiments_markers/[expName]/marker.mp4.

Run directly in command line

1. Run the experiment

   stdbuf -o0 ./build/main \
       press_example \
       true \
       true \
       1e-20 \
       10 \
       true \
       resources/gel/gelslim-gel-l_0.1/gelslim-gel-l_0.1.msh \
       resources/press_obj/Wave_8134.obj \
       resources/gel/gelslim-gel-l_0.1/isBC_gelslim-gel-l_0.1.json \
       0 \
       press \
       6 \
       1 \
       25 \
       0 \
       25 \
       0.01 \
       > output/logs/press_example.log

2. Postprocess and obtain marker displacements (same as above)

Run custom experiments

• Run experiment with custom configurations

  Modify scripts/config.py and run_experiments.py, or just run command following

  stdbuf -o0 ./build/main \
      [expName] \
      [enableGui] \
      [enableCGSolver] \
      [cgRel] \
      [PNCap] \
      [enableInversionPrevention] \
      [gelPth] \
      [objPth] \
      [isBCPth] \
      [fricMu] \
      [moveType] \
      [pressSteps] \
      [pressDepth] \
      [pressVel] \
      [taskSteps] \
      [moveVel] \
      [dt] \
      > [logPth]

• Postprocess custom tactile sensor gel

  1. Prepare your tactile sensor gel mesh and boundary condition file.
     • The gel mesh should be tetrahedral and in .msh format. The length should be in milimeters.
     • The boundary condition file should be in .json format. It is a list of binary numbers, where 1 indicates the corresponding node is fixed, and 0 indicates the node is free to move.
  2. Take gel_0.obj after experiment as surface mesh(recommended), or extract it mannually.
  3. Generate the triangular barycentric weights on the surface mesh for markers, and save as marker_bc_ws.pkl in the same directory as your .msh gel mesh. You may refer to process/gen_markers.py.
  4. Postprocess as usual.

• Custom tasks

  Refer to main.cu to implement your own tasks.

  • Program entrance
    • Modify arguments in main()
  • Task implementation (in prepare())
    • Modify configuration to be dumped in config
    • Modify solver settings
    • Modify move lambda to implement your own task (usually changing the target state a bit each step)

Cite this work

If you use TacIPC in your work, please cite us.

@article{tacipc2024,
  title={TacIPC: Intersection-and Inversion-free FEM-based Elastomer Simulation For Optical Tactile Sensors},
  author={Du, Wenxin and Xu, Wenqiang and Ren, Jieji and Yu, Zhenjun and Lu, Cewu},
  journal={IEEE Robotics and Automation Letters},
  year={2024},
  publisher={IEEE}
}