[StabilizerTask] Fix frames of forces and velocities in foot force difference control #290
Conversation
…ce difference control. This should fix jrl-umi3218#289
| @@ -1142,15 +1142,19 @@ void StabilizerTask::updateFootForceDifferenceControl() | |||
| return; | |||
| } | |||
|
|
|||
| sva::PTransformd T_0_L(leftFootTask->surfacePose().rotation()); | |||
| sva::PTransformd T_0_R(rightFootTask->surfacePose().rotation()); | |||
There was a problem hiding this comment.
I start to consider these lines should be
sva::PTransformd T_0_L = leftFootTask->surfacePose();
sva::PTransformd T_0_R = rightFootTask->surfacePose();
for wrench transformation although we used only the rotation in the similar case https://github.com/arntanguy/lipm_walking_controller/pull/2/files#diff-86694886f91b285c323e710f1b64ec1c0f5ab403d0ee748e112030f766fcae11R139
I'm not confident, so could anyone give some idea?
There was a problem hiding this comment.
I reconsidered and came to the conclusion that it would be OK to leave it as it is ("as is" in the merged PR).
For wrench, only force is used (moment is not used), so the position of the transformation is not important.
For velocity, the angular velocity is set to zero as shown below, which implicitly assumes a frame whose position is identical to the surface and whose orientation is identical to the inertial frame. In such cases, it is correct to apply rotation-only transformations.
https://github.com/jrl-umi3218/mc_rtc/pull/290/files/72f45b20f3766351b296e37b734b65113184baeb#diff-be940a972e16470195608839479a680e1f8aba715c2e5f4c62cdaa4666be420eR1169-R1170
Changes --- - [general] mc_rtc now uses C++17 (#297) Added --- - [mc_control] Publish joint sensors in ROS (#293) - [mc_rbdyn] Add a non throwing version of ZMP computations (#292) - [mc_rbdyn] Add const accessor for grippers (#298) - [mc_rtc] Add gui::Polyhedron element (#296) - [mc_tasks] DCM derivative gains are now 2D (#295) Fixes --- - [mc_control] Fix gripper initialization issue (#291) - [mc_control] Fix assumption around robot.hasJoint - [mc_tasks] Set roll and pitch target of torso to 0 in StabilizerTask (#175) - [mc_tasks] Fix foot force difference control issue in StabilizerTask (#290)
This should fix #289
I've confirmed that the JVRC1 robot can walk in lipm_walking_controller in Choreonoid with this patch.
However, there is no explicit confirmation that this fixes the bug #289 effectively. (It is also not clear how to identify bugs from the robot's behavior and logs in the first place.)