Trajectory Planner using Ruckig Lib

zmrdko wrote:

How about collect $ upon delivery then?
 

I think a bounty like that could be a good idea.

I really understand most people here work in the code and documentation as a hobby, and get great happiness when things work and can be shared with the community,obviously put a money in the middle convert a hobby in stress, frustration, regrets, claims, etc. and no body want these factors in its life.
In the other hand most people here is agree jerk control will take LinuxCNC to the next level, but this is a complex thing to achieve.I think this should be a project that need resources maybe many of you are familiar with KICAD a PCB software, somehow the get resources to make huge improvements in the software,they hire programmers dedicated to creating specific features and fix specific issues.
I think cofounding is a good idea, but get the money would be the easy part, then we need to create a project, clearly define what do we want,hire the programmer or programmers, and continuously monitor the progress and effectiveness of the work.I really don’t know any programmer that can have experience on this specific field.Bottom line, I agree to donate and support in case of the community want to go this way

None of this really works well in practice. it needs (again) someone like tormach to just pay someone to do it for THEIR needs, and we get the spoils.

Otherwise, we can sit and watch and wait and just be happy.

interesting, so basically some one has to take the initiative collect the money, hire the programmer, follow the project and then release the results to the community. 

Hi,

Last day's i was working again on the scurve implementation.

I have a working scurve program. But look ahead has to be done coming time.
It uses Grbl's corner algoritme. And uses a grotius scurve for the tp, using end velocity's based on the Grbl's output.

The Grbl's deviation or delta factor is set to 0.001 for this video. If using 0.0001 corner speeds are more reduced.
Mention no look ahead is active.

Hi,

Yesterday i started to config a production machine to run on linux, running the scurve.

I had some troubles off course.

Some off my findings :

- Linuxcnc run's In simulation mode when i code. Then everything runs ok. However on the production machine the tpmod_scurve.so would not load properly. I could not find the cmake problem in this time span.

- So to overcome this problem, i used the original linuxcnc make system. I added the files to the makefile and tried to
compile. But one off my files had a cpp extension. So i tried to modify the makefile to filter out this filetype, but it would
not succeed in a reasonable time span.

- In the end i edited the .cpp file to .c The tpmod_scurve.so was compiled and linked ok. The simulation runs ok.

Now i hope tonight to run this on the production machine. So this test involves the jogging component only.
When this run's ok. I will go on the get the trajectory planner fertig.




 

Hi,

Today i did the scurve jogging test on a production machine.

This machine is a XXYZAB router with steppers, driven by a dual parport.
The machine run's about 12000mm/min max in the video's. It's an old machine,
so you can hear some drive sounds.

For the test i used a glass filled with H2O fluid. We can clearly see when things go smooth or not.

Using acceleration 325
Video 1 -> C0400 is trapezium curve.
Video 2 -> C0401 is scurve.

Using acceleration 1000 :
Video 3 -> C0402 is trapezium curve. ( Here the machine breaks down at low jog speeds )
Video 4 -> C0403 is scurve. ( The machine run's up to 12000mm/min, no problems )

Video 1 trapezium:

Video 2 scurve:

Video 3 trapezium, breaks down at acc 1000:

Video 4 scurve acc 1000 maxvel 12000


For me the first jogging test was succesfull. I could clearly see the scurve is able to use higher acceleration
profile. Also the H20 fluid keeps quite steady with scurve. The downside is when using scurve jogging you can
experience some kind jog delay. This is i think typicall for scurve, when acc up, and button is released the acc
curve has first to go down to acc zero. From there the jog stop curve comes in.

So far so good.



 

That is an impressive difference!

awesome!

Hi,

That is an impressive difference!
Yes i agree.

awesome!
I have a new ethercat production machine, this machine is a candidate to test the scurve on.

Project update:

I finally tracked down the cmake compile error, with makes linuxcnc runs in non-realtime mode.
Now i am able to us the compiled cmake code in realtime.

Lcnc startup output now ok for loading scurve trajectory planner:
linuxcnc TPMOD=tpmod_scurve_grotius HOMEMOD=homemod EMCMOT=motmod
Note: Using POSIX realtime

This is solved it in the cmakefile of the component, using a static link to the lcnc/lib dir.
# Link to libraries.
target_link_libraries(${PROJECT_NAME} PRIVATE linuxcnchal -Wl,-rpath=/home/user/linuxcnc/lib/)

Now this is not a beauty thing, i will make a better solution later on.