Replies: 1 comment 1 reply
-
In this case the
This will impact the solver behavior because it results in an equation being added to the solver. I would expect different behavior when changing the value. I usually do not set this value to a vary large number because currently I believe this only modeled as a linear equation and not a quadratic. This is most to prevent bouncing around at the threshold which speeds up convergence. I usually only set this to a few centimeters.
This is a known limitation for you environment. Where joint 5 and 6 will provide gradients to move away from the robots origin and then joint 4 will provide a gradient to move toward the robots origin. This will prevent it from solving, which is why when you only the closest is provided it succeeds. |
Beta Was this translation helpful? Give feedback.
-
Using the same setup from #263, I am also running into some strange behavior with the contact test type and collision checking:
Case 1
Result: success
case_1.mp4
Case 2
Result: the optimization somehow succeeds, but clearly some collision constraints are violated, as seen in the video and the
trajopt
logcase_6.mp4
Case 3
Result: optimization now fails because the collision constraints are not met and the
safety_margin_buffer = 0.0
Questions
safety_margin_buffer
seem to have something to do with how the collision constraint is evaluated?ALL
tend to fail for a case like this?Beta Was this translation helpful? Give feedback.
All reactions