I post this piece in CNC section, as it deals with 5-axis setup but with FDM in mind.

Issue

I have a simple 5-axis FDM setup, 3-axis X,Y,Z plus Z-rotation and tilt of the FDM nozzle; and due to the actual setup I need to do an inverse kinematic (IK) calculation. I can add this IK to the firmware of Duet itself, but I also can post-process G-code from tool coordinates to machine coordinates (I just did this as an experiment).

Advantages:

• no change in firmware (no recompiling)
• simple profiling (change setup needs no change in firmware required), each setup/machine has its own profile (a JSON file describing the inv kinematics)
• simple calibration of actual (physical) offsets and angles

Disadvantages:

• machine dependent G-code
• another step to convert G-code
• risk of doing stupid things (e.g. running tool into the bed) as firmware doesn't prevent it

To any experienced CNC machinist and multi-axis wizard, what are your thoughts? I don't know how CNC industry in general deals tool vs machine coords/g-code, and I like to get to know some good practices. Thanks.

If you like to see the actual setup https://xyzdims.com/2021/02/08/3d-printing-penta-axis-pax-5-axis-printing-option/ (work in progress)

@Miss-Rebekah @droftarts thanks for the feedback.

I realize in CNC world the CAM software does the heavy lifting, whereas with 5-axis FDM in its infancy I'm currently bootstrap all myself: planar/non-planar slicing (works) and now I reach the decision whether to implement inverse kinemantic (tooltip coord vs machine coord) in the firmware or leave to the post-processor (prototype exists already).

I might end up doing both approaches and then see which one is more usable in real life.

@joergs5 I've got my kinematics numerically working with the few helpers based on FixedMatrix<floatc_t,4,4> I mentioned (not yet optimized or so).

Still some stuff to sort out regarding LimitSpeedAndAcceleration() to coordinate different speeds of my axes (they differ quite a lot in speed).

@joergs5 I just tried the library you mentioned, but I get many errors when including it, it struggles with ios dependency (within matrix_library/matrix.h by include of<iterator> and <complex> (I didn't find an easy way to resolve it).

@dc42 @JoergS5 I gonna code some matrix helpers for my use case, once things run well and reliable I get back to you and you can look at my code and consider to include something in RRFLibrary/Math.

PS: I spent hours chasing errors reported under Eclipse (Linux), but console reported 0 errors - once I ran Index > Rebuild, the errors under "Problems" disappeared

I'm currently coding for Penta Axis (PAX) 5-axis FDM option https://xyzdims.com/2021/02/08/3d-printing-penta-axis-pax-5-axis-printing-option/ on Movement/Kinematics/PAXKinematics.{cpp,h} aka M669 K14 @dc42 and I need some guidance to implement some inverse kinematics needed for MotorStepsToCartesian() & CartesianToMotorSteps(), something like:

matrix4x4 m2 = mtranslate([xoff,yoff,zoff]);
matrix4x4 m1 = mrotate([xrot,0,0]);
matrix4x4 m0 = mrotate([0,0,zrot]);
matrix4x4 m = minvert(m2*m1*m0);  

x = m.elem[0][3], y = m.elem[1][3], z = m.elem[2][3];   // extracting translation part again

It's a bit more complex, but in a nutshell I need a sequence of rotate/translate in different orders, and inverse at the end, and then translation vector extraction and thought doing this more high level with 4x4 matrices.

I'm not versed in C++ and Eclipse but I know C sufficiently, but it's 20+ years gone by since (Perl/Python these days).

It's not clear to me whether basic matrix ops are available in RRF, and which library is available and/or used, it seams Math/Matrix.h provides mainly structural functionality (create & set), and inversion of matrices (GaussJordan reduction), but not multiplication of matrices, correct?

What's the best way to implement mtranslate(), mrotate() and minvert() in RRF?

Cc: @JoergS5 (author of FiveAxisRobotKinematics.cpp)

@roiki11 You might look at my blog post https://xyzdims.com/2021/04/19/3d-printing-sub-volume-segmenting-non-planar-slicing/ regarding 4- and 5-axis FDM and non-planar slicing.

@Miss-Rebekah @droftarts thanks for the feedback.

I realize in CNC world the CAM software does the heavy lifting, whereas with 5-axis FDM in its infancy I'm currently bootstrap all myself: planar/non-planar slicing (works) and now I reach the decision whether to implement inverse kinemantic (tooltip coord vs machine coord) in the firmware or leave to the post-processor (prototype exists already).

I might end up doing both approaches and then see which one is more usable in real life.

I post this piece in CNC section, as it deals with 5-axis setup but with FDM in mind.

Issue

I have a simple 5-axis FDM setup, 3-axis X,Y,Z plus Z-rotation and tilt of the FDM nozzle; and due to the actual setup I need to do an inverse kinematic (IK) calculation. I can add this IK to the firmware of Duet itself, but I also can post-process G-code from tool coordinates to machine coordinates (I just did this as an experiment).

Advantages:

• no change in firmware (no recompiling)
• simple profiling (change setup needs no change in firmware required), each setup/machine has its own profile (a JSON file describing the inv kinematics)
• simple calibration of actual (physical) offsets and angles

Disadvantages:

• machine dependent G-code
• another step to convert G-code
• risk of doing stupid things (e.g. running tool into the bed) as firmware doesn't prevent it

To any experienced CNC machinist and multi-axis wizard, what are your thoughts? I don't know how CNC industry in general deals tool vs machine coords/g-code, and I like to get to know some good practices. Thanks.

If you like to see the actual setup https://xyzdims.com/2021/02/08/3d-printing-penta-axis-pax-5-axis-printing-option/ (work in progress)