Adds IMU projected_gravity_b and IMU computation speed optimizations #2512
Conversation
jtigue-bdai
requested review from
jsmith-bdai,
pascal-roth,
kellyguo11 and
Mayankm96
as code owners
|
Can we keep quat rotate and quat apply changes to another MR? Otherwise it is getting burried here under "IMU changes" which isn't correct. |
I can, I was mainly trying to utilize the quat_apply_inverse function which was introduced in the other PR. If we want to get this in sooner I can switch to quat_apply(quat_inv(quat)). |
jtigue-bdai
force-pushed
the
jat/feats/imu_improvements
branch
from
e0b096e to
8067b49
Compare
| # default to all sensors | ||
| if len(env_ids) == self._num_envs: | ||
| env_ids = slice(None) | ||
| # obtain the poses of the sensors | ||
| pos_w, quat_w = self._view.get_transforms()[env_ids].split([3, 4], dim=-1) | ||
| quat_w = math_utils.convert_quat(quat_w, to="wxyz") | ||
| quat_w = quat_w.roll(1, dims=-1) |
There was a problem hiding this comment.
is this because convert_quat has more overhead?
There was a problem hiding this comment.
yeah there is a little bit more logic. I was trying to squeeze as much as possible out.
There was a problem hiding this comment.
I see, how significant is the difference? wonder if it makes sense to make a more lightweight version of convert_quat maybe just for torch tensors?
There was a problem hiding this comment.
I don't think it was that much different, but I don't have the numbers for that.
kellyguo11
changed the title
Description
This PR adds projected_gravity_b data stream as well as speeding up computation by during batch rotating of all data streams to base frame in one computation rather than each data stream separately.
This PR is dependant on PR 2129 and should not be merged until that PR is merged.
Fixes # (issue)
Type of change
Checklist
pre-commitchecks with./isaaclab.sh --formatconfig/extension.tomlfileCONTRIBUTORS.mdor my name already exists there