scurve trajectory planner

Hi,

I have tried another solver: "libgsl".     ( sudo apt-get install libgsl-dev )   Using the GPU. ..

This runs way better then libdogleg. Was a luckie shot.

In this solver bigger distances are ok.
Input :


interpolated endpoint result : x: 520.049999 y: 266.500001 z: 125.500000
Converged after 370 iterations wtf? 

Now i get more convinced finally.

Moved to another PC, installed 'libdogleg-dev' and I seem to get a little further but not much:

 

Anyway, I'm afraid there is not much I can think of to help as I'm really starting a zero with these dogleg solvers.

Hi Arciera,

use libgsl example. Remove all libdogleg related files and deps.

yes that looks much better and compiles for me too.

@Arciera,

Today i started coding the clothoid compound fit using the gsl solver. It worked right away.
I think because yesterday we did it succesfully for just one clothoid.

This is really a milestone for us !!!

What we see here is a 3d clothoid spline, build from 4 individual clothoids, given start, end posture.

Given the abstract :
"Space corner smoothing of CNC machine tools through developing 3D general clothoid"
by Qun-Bao Xiao, Min Wan, Yang Liu, Xue-Bin Qin, and Wei-Hong Zhang.
Published in Robotics and Computer Integrated Manufacturing, Volume 64, 2020.


However if you look closely at the individual clothoid curve transitions there is a tiny change in curvature.
Somewhere in the code is a buggybug.    and or alien obstruction.
I think in eq45 where the constrains of the joints are calculated must be something signed to "-" ?

 

This is really a milestone for us !!!

Indeed! Congratulations.

However if you look closely at the individual clothoid curve transitions there is a tiny change in curvature.

Is eq 45 in the loop yet?

I can't quite tell from your repo so you may already have but just in case:
Remember that we need to run eq 45 with the gamma11, gamma21 and s1 values found with dogleg. With the 6 resulting gamma values it should then satisfy G1, G2 and G3 continuity.

 

I don't think these gamma values look right:

Actually the problem might be in the dogleg algo, seems odd to me that gamma11 and gamma21 are zero.

Solution found: gamma10 = 0.000000, gamma20 = -0.000000, s1 = 198.587624
 

Could it be that we are only solving for variable s1 instead of [gamma11, gamma21, s1]?

@Arciera,

I made a test : clothoid_compound_no_fit();

This visualises the problem quite good. This example uses no fit for endpoint. So the clothoid compound
behaves as a natural attached curve.

eq44 is reviewed and i added the for loop check calculations as in the paper.
However the tresholds results there are ok.
eq45, i can not find any typo's.

eq45 is sub off eq44_G123. Yes eq45 is in the loop.

With the 6 resulting gamma values it should then satisfy G1, G2 and G3 continuity.
Yes. eq45 is called for xy and for z.

Then yi1 is called like : y11 for z
Then yi1 is called like : y21 for xy.

This to avoid code duplication.