Kinematic Data Extraction Guide.

[matangpanchal]

Can anyone please give some hint to derive the kinematic model data such as below:

Vec3F base_offset(-32.5f, -32.5f, -32.5f);
arm_length = 73.8f;
rotor_radius = 15.0f;
ee_attachment_points[0] = Vec3F(-0.5f, -14.5f, 2.0f);
ee_attachment_points[1] = Vec3F(2.0f, -0.5f, -14.5f);
ee_attachment_points[2] = Vec3F(-14.5f, 2.0f, -0.5f);

// set transformation based on CAD model
actuator_to_base[0].rotation = QuaternionF::from_axis_angle(Vec3F(0.0f, 0.0f, 1.0f), 90.0fD2R);
actuator_to_base[0].translation = Vec3F(-42.0f, 0.5f, 32.0f)+base_offset;

actuator_to_base[1].rotation = QuaternionF::from_axis_angle(Vec3F(1.0f, 0.0f, 1.0f), 180.0fD2R);
actuator_to_base[1].translation = Vec3F(32.0f, -42.0f, 0.5f)+base_offset;

actuator_to_base[2].rotation = QuaternionF::from_axis_angle(Vec3F(-1.0f, 0.0f, 0.0f), 90.0fD2R);
actuator_to_base[2].translation = Vec3F(0.5f, 32.0f, -42.0f)+base_offset;

rotor_angle_offset[0] = 46.2fConstants::DEG2RAD;
rotor_angle_offset[1] = 46.2fConstants::DEG2RAD;
rotor_angle_offset[2] = 46.2fConstants::DEG2RAD;

We have made some changes in the dimensions in our design , so We need to derive this info from the CAD model.

[0x23]

Below is a slightly more commented version for the (not yet release) 4.0 design. I also recommend watching the part about the kinematic model from the video. It shows how the actuator coordinate system is defined (the one an the motor axis).

I hope this helps a bit. I currently dont have the time to document the derivation from the CAD model but its on the TODO list.

KinematicModel_Delta3D::KinematicModel_Delta3D() {
  const float D2R = Constants::DEG2RAD;

  // location of base origin in endeffector coordinate system (ee in neutral position)
  Vec3F base_offset(-47.5f, -47.5f, -47.5f);
  
  // distance between the two spehere centers
  arm_length = 36.25f*2;

  // distance from arm attachment points to rotor axis
  rotor_radius = 15.0f;

  // midpoint between the two attachment spheres on the endeffector in EE coordinate system
  ee_attachment_points[0] = Vec3F(3.54f, -11.5f, 4.5f);
  ee_attachment_points[1] = Vec3F(4.5f, 3.54f, -11.5f);
  ee_attachment_points[2] = Vec3F(-11.5f, 4.5f, 3.54f);

  // set actuator transformations based on CAD model
  actuator_to_base[0].rotation = QuaternionF::from_axis_angle(Vec3F(0.0f, 0.0f, 1.0f), 90.0f*D2R);
  actuator_to_base[0].translation = Vec3F(-21.5f, 21.0f, 52.0f)+base_offset;

  actuator_to_base[1].rotation = QuaternionF::from_axis_angle(Vec3F(1.0f, 0.0f, 1.0f), 180.0f*D2R);
  actuator_to_base[1].translation = Vec3F(52.0f, -21.5f, 21.0f)+base_offset;

  actuator_to_base[2].rotation = QuaternionF::from_axis_angle(Vec3F(-1.0f, 0.0f, 0.0f), 90.0f*D2R);
  actuator_to_base[2].translation = Vec3F(21.0f, 52.0f, -21.5f)+base_offset;
  
  // angle offset from home positionto center position of the rotor
  rotor_angle_offset[0] = 42.0f*Constants::DEG2RAD;
  rotor_angle_offset[1] = 42.0f*Constants::DEG2RAD;
  rotor_angle_offset[2] = 42.0f*Constants::DEG2RAD;

  for(int i=0; i<3; i++)
    base_to_actuator[i] = actuator_to_base[i].inverse();
}

[matangpanchal]

Sir, I am very thankful for your response. Data in this code is different than the data in the Github sourcecode file, So Cad files or assembly files for these dimensions are different then the provided on Github Repo ? Or, If I will measure distances as per the comments then I will get these data only ? This matching I want to validate my understanding only.
Thanks Again.

[0x23]

The data is different because its from a new design (v4) which has not been released. Ignore the actual numbers. I just provided the code because of its comments in the hope it may help you to better understand whats going on.

[0x23]

The ee_attachment_points are the midpoints between the two attachment balls in the coordinate system of the endeffector (not the global coordinate system!):

[0x23]

The actuator coordinate system is defined like this:

it is located on the motor axis and its Z coordinate is again in the center between the two balljoints:

actuator_to_base is then the transformation between each of the actuator coordinates systems and the global device coordinate system

[matangpanchal]

Thanks for more details, I will try figuring out further.

[matangpanchal]

Based on your input and with some trial and error I have derived nearby co-ordinates for my hardware cad Files. I have documented how to derive required values using CAD modules according to my understanding, which is attached with this post. You may refer and use it if its appropriate.
Further I have a question regarding Calibration.

I have use your design reference to make similar HW only, otherwise I am not using the firmware and otherstuffs, I am trying to integrate this HW with my software and till some extents I got some success as well.
Now I am looking for Calibrating the HW. Do you have any video or reference document to calibrate the HW with Software to mitigate Manufacturing and assembly errors ? If I get some reference I can make similar system for error mapping and compensation.

Micromanipulator Kinematic CAD Mapping.pdf

[0x23]

The way you derived the values looks correct to me. Thank you for providing a documentation for others. Currently only the encoder values are calibrated to the motor position given an electric field angle (done by the firmware). This is not ideal, since rotor position will not follow electric field angle exactly (e.g. due to cogging). I plan to improve on this by using one of the magnetic encoders directly on the motor axis as reference.

Calibration of geometric errors and error mapping is currently not done. Main problem here is the 6-DOF measurement of the endeffector with sufficient accuracy over a 20x20x20mm range. If you have any insights on how this can be achieve, please share...