c2 smooth velocity profile

Hi,
thank you for your efforts regarding a new trajectory planner.
I've tried your branch with the smooth planner. I've noticed a strange behavior of very large accelerations. I'm posting an example GCODE file and screenshot of halscope.

I'd like to push this forward. Can you please suggest me how to start debugging?
Best

The TP probably replaces the sharp corner with a small circle, and entering/exiting the circle causes big acceleration spikes on the joints, my approach doesn't get rid of those. But this looks like some target is not reached or it is overshooting and some correction is applied before starting the next segment. Suppose you need to come to a stop, the smooth profile decelerates such that the target point is reached with velocity 0 and acceleration 0, but if there is some slight inaccuracy or change in target or bug / wrong assumption then it is possible velocity 0 is reached before the target and then you will see a spike (at the end of a segment, commanded position is set to target position).

Off the top of my head, there should be some debug pins in the TP, try adding those to halscope.

I will take a look in the evening.

Thanks, you we're right. The strange acceleration spikes are caused by the blending mode. If blending mode is set to G61 (exact path), the program just works fine. I attached the outputs from halscope.

With Blending enabled (G64 P0.5):


Without blending enabled (G61):

 

I'm curious to see how this performs - thanks for the heads up @hyroid7 on the blend mode

So this works quite well, I'm really impressed.

I am an adventurous type and have a good e-stop system so I connected it to my 3 axis router (I've crashed him enough without the help of bugs so I'm not very worried) and it seems to work well so far. Smooth

Are there any parameters that are adjustable that I can play with both in sim and on the machine? I would like to figure out how to find a balance between the G61 smoothness and the G64 with blending jerkiness to keep feedrates up for roughing, but for finishing passes G61 works amazing.

OK, so I've potentially found my first edge case.

When jogging into the tool length sensor, it properly stops when the sensor trips, as expected.

However, when running a subroutine (github.com/TooTall18T/tool_length_probe) the motion continues regardless of the probe touch. I didn't wait to see if the overtravel stop (connected to software e-stop) on the length sensor worked properly or not, although my assumption is that it would (as manually halting using F2 key stopped its progress immediately).