Trajectory Planner using Ruckig Lib

Thanks Grotius


Yes the multi axis stuff is where it would be great to have a decent lookahead

And we would really appreciate it

Hi,

If i run this in a 9 axis config, it blends the corner, and therm condition=2. Wich means TC_TERM_COND_PARABOLIC.
It has a 4th axis a.



Maybe i need a sample file where the motion stops each segment.

you can see a couple of example trajectories here -  rotary toolpath .
simpleTP with parabolic blending and 1 line look ahead and there is a reason for slow velocity - it never reaches target speed on them...

No problem with large motions(x0a0 -> x100a100) - all the problem is that any multi-axis trajectory = polyline of many microscopic motions(I attached two examples with screenshots at the link). Simple parabolic blending on each of line is not enough, you need global blending with tolerance... (what Rob Ellenberg's planner does for XYZ axes).

Grotius

I have a large 4axis program I can attach here if you like.

As others have said the issues arise when posting a lot of high accuracy gcode any move that includes the 4th axis (A axis) means that the look ahead reverts to 1 line.

Rather than something like 1000 lines ahead

@hmnijp,
Thanks for your contribution.

@Beckvill,
Please post the file, than i can save it to investegate later on.

For the question to enable 4th axis without disabling look ahead, i thought of making a simple solution.
This solution should be fine for 3d printers using a 4th axis for a extruder. The idea was to interpolate
the 4th axis given the current gcode line progress 0-100%.
And try to add it up to the fillet. 

For this i have to figur out the original trajectory planners workflow for fillets and make a intrusion somewhere.
This will take more then one day to investegate. So if this is a bad idea, i better do something else.

 

Hi,

Made a sample video of the progress so far for hal-core-2.0.

The gui sends the jog commands to the hal-component "state_machine.so" via shared memory.
Jogging is done by using scurve.
Motion speeds, acc, jerk, etc, is saved within the program settings tab. No ini file is needed so far.
Qt has a nice save settings framework, wich works ok.

In the state_machine.so component, all info is there to write the trajectory planner coming time.

So far the code is compact.

was doing a simple little part with some very common deburing operation so throught id share that part of the code for you here to test with,
i have attached 2 sample G code File, both posted from fusion360 for multi axis deburing,
is posted in Inverse feed mode (G93)
feed rate is quite low due to linuxcnc inability to process multi axis code smoothly so hopefuly will aid in ability to soon run much smooth and faster

use as you wish, I hope it helps with testing out the planner etc on multi axis side of things

if intrested here is a video of what it should do when simulated.


setup wise its posted for a machine of a 3axis + B & C rotary indexer, C rotates around Z , B Around Y
DEBURING OPERATION 5AXIS 2 Contains B & C moves
DEBURING OPERATION 5AXIS  only C Moves in the feed contour

Hi Grotius,
I have been trying to find out the physical meaning of curvature and how to imagin G2 continuity in higher dimensions. I have not been having much success in that endavour.
This recently published paper I have found claims to use splines for filleting in 5 axis: A Novel Local Smoothing Method for Five-Axis
Machining With Time-Synchronization
Feedrate Scheduling ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=9085380

I was hoping for some help to find some references on how to do filleting in higher dimensions to reduce the tangential jerk.

-automata

One possibility that comes to mind is to
1. smooth the tool position trajectory (ie XYZ) as Grotius has been doing
2. create a second 3d trajectory that is offset from the tool position trajectory by a fixed distance along the tool orientation vector
3. smooth this second trajectory using the same method as in step 1
4. calculate the smoothed tool orientation vectors
5. calculate the smoothed orientation angles (ABC) from the smoothed tool orientation vectors.


I found a paper that may describe the method above but I do not have access to the full text:
www.researchgate.net/figure/3D-clothoid-...ector_fig3_361356637

@Automata,

Nice document.
There is a video link from andy, in the video they show curvature.

@Arciera,
I think also a graphical representation of the abc optimalistion will be much easyer to do. For sure in opencascade.
If i draw the tool orientation vector lenght:10mm at each gcode line. Or at each known position.

Then i connect these tool orientation's at the top with a poly line.
Then create fillets inbetween the polylines? Then set tool orientation vector to touch the fillets?

Maybe a scetch would help to verify the steps.