introvoyz045's picture
Migrated from GitHub
ef17c16 verified
|
Raw History Blame Contribute Delete
2.98 kB

Using cisstRobotPython to replicate the forward and inverse kinematics of the dVRK

Compilation and environment

By default the Python wrapper for the cisst libraries are compiled when using catkin build. If you compiled the code using CMake directly, you need to make sure the SWIG_PYTHON flags are properly set when compiling cisst.

To set your Python path so Python can find cisstRobotPython, you can use the script cisstvars.sh. When using catkin:

source ~/catkin_ws/devel/cisstvars.sh

We provide a basic example. To run it, make sure you are in this directory and type:

python3 example.py

Main differences with the dVRK C++

  • cisstRobotPython can't load .json files directly, instead you will have to use .rob files (some examples can be found in directory share/deprecated)
  • For the PSM, the dVRK 2.0 uses two DH files, one for the first 3 joints and one for the instrument. This can be achieved in Python by loading two separate .rob files consecutively. The .rob files don't exist so you will have to create your own from the psm.json and the <instrument>.json files (see directories share/tool and share/)
  • base-frame can't be defined in .rob file but you can add one in Python (see below)
  • Starting with the dVRK 2.0, the ECM inverse kinematics uses a closed form computation and this is not available through Python

Troubleshooting and tips

  • Failed to import cisstRobotPython, error looks like:

    ImportError: dynamic module does not define init function (init_cisstRobotPython)
    

    This can occur if the Python wrappers were compiled for Python 3 but you're using Python 2. You can either change your Python interpreter or recompile the wrappers for Python 2. To do so, run CMake in the cisst build tree (ccmake ~/catkin_ws/build/cisst and change all the references to Python 3 to Python 2). Then re-compile your catkin workspace with catkin build.

  • Check the return value of LoadRobot.

    r = cisstRobotPython.robManipulator()
    result = r.LoadRobot('/home/cstar/catkin_ws/src/cisst-saw/sawIntuitiveResearchKit/share/deprecated/dvpsm.rob')
    # result should be 0 if successfully loaded
    # one can also make sure the number of links is correct
    print(r.links.size())
    
  • To modify the base frame, assign a 4x4 matrix to r.Rtw0

  • To add a tool tip transform, create a dummy manipulator and attach it to the existing manipulator:

    tt = robManipulator()
    tt.Rtw0 = tooltip
    r.Attach(tt)
    
  • Accessing and modifying DH parameters:

    l2 = r.links[2]
    k2 = l2.GetKinematics()
    k2offset = k2.PositionOffset()
    k2.SetPositionOffset(k2offset + some_noise)
    
  • Calculating the Jacobian:

    jp = np.zeros(6) # joint angles
    J_body = np.zeros(shape=(6,6), dtype=np.double) # placeholder
    r.JacobianBody(jp, J_body)
    print(J_body)