Move Trajectory planner from real time to non-real time

hi guys，
I wonder if it is possible for us to move the trajectory planner from real time to non real time.
1.TP is  a time-consuming part,It should not be too complex, otherwise it will affect the stability of real-time cycle；
2.we can not perform complex path smoothing algorithm,such as B-spline Compression smoothing (siemens fanuc),with current structure;

If we move TP to non real time module, we can do more complex smoothing calculation.

Has anyone tried this before?

I'm going to try this job. Do you have any suggestions?

 

How will you handle such things as override and spindle sync?

and eoffsets..

I think the LinuxCNC derivative "OpenCN" does something similar
(it seems to trade off real-time planning to simplify things like jerk control)

Yes,I've checked machinekit.
Machinekit uses almost the same trajectory planning method as linuxcnc.
Machinekit calculates the original trajectory equation \ create blend-arc path between TCs \ optimize TCs finalvel in tpAddLine() /tpAddCircle().
I know that because I developed a branch from machinekit. Perhaps I will introduce my previous work and research later.

andypugh wrote:

Have you checked to see if machinekit have already done this?

Yes,I've checked machinekit.   
Machinekit uses almost the same trajectory planning method as linuxcnc.
Machinekit calculates the original trajectory equation \ create blend-arc path between TCs \ optimize TCs finalvel in tpAddLine() /tpAddCircle().
I know that because I developed a branch from machinekit.  Perhaps  I will introduce my previous work and research later.

PCW wrote:

I think the LinuxCNC derivative "OpenCN" does something similar
(it seems to trade off real-time planning to simplify things like jerk control)

Thanks PCW !  Very useful tips！
I read "OpenCN" project doc and their youtube channel,  They did put the trajectory calculation and speed optimization modules in non real time module. They use matlab to program the parts of trajectory planning and speed optimization, and then use the tools provided by MATLAB to convert Matlab code into C code. Very brilliant.
 
 
Muti-cpu framework：cpu#0  cpu#2,3  - non real time,   cpu#1 real time.

However, after knowing the G code they can support (the following fig: a lot of g-code are not supported), I feel that this project(openCN) is more inclined to academic experiments than engineering applications. 



Although this project seems unable to be used directly, I will continue to study it，maybe it can bring some new inspiration !  Thanks a lot anyway!