Twist controller extensions #110
Conversation
MaChristoph
commented
Jul 26, 2016
MaChristoph
commented
- Added a new constraint for Joint Singularity Avoidance (JSA). This doesn't work right now and we're trying to change this constraint to maximize the manipulability rate and find a better gain for the new introduced singularity avoidance algorithm
- Added new singularity avoidance algorithm with 3 new parameters in the RQT dynamic reconfigure (damping_delta, damping_gain, damping_slope). This behaves in singular positions better and calculates for each joint its own damping value lambda. (cp. the old one calculated one damping value for all joints)
- Added base_active mode for stack_of_tasks controller. I wasn't able to verify this 100 % since the base_active doesn't behave as I want to ( extension_ratio needs to be derived dynamically ). The new singularity avoidance algorithm with manipulability optimization might be the right way to achieve this goal. We want to test this until Friday.
…cle_distance_moveit_integration
…ss).. we might change this one to maximize the manipulability rate to find a better value for the damping_gain which we introduced to the pseudo inverse calculations. The main change in the code is the additional constriant.
…oller_extensions2 Conflicts: cob_twist_controller/CMakeLists.txt
|
I can't really review this on the phone...but few things I found:
|
|
Also, please document the literature you used! Where did you find those concepts/the formulae therefore...please post a link! In case one of the features is not working correctly, open a new issue respectively referencing this PR, exactly describing the problems, the ways you tested it and how to reproduce... I'll be checking in tomorrow again ;-) P.S.: |
| @@ -93,6 +93,7 @@ std::string chain_base_link_; | |||
| marker_vector.id = 42; | |||
| marker_vector.header.frame_id = chain_base_link_; | |||
|
|
|||
|
|
|||
|
Unfortunately, there are still a few things I don't like about this PR... So...without someone finalizing, documenting and testing this feature...I rather would not merge this PR... |
