Proposal
Extend FrameTransformer so it can optionally report linear and angular velocity, and optionally linear and angular acceleration, for the source and target frames, not only pose. Data would be exposed in both world and source frames (mirroring existing *_w / *_source pose fields), using the same rigid-body view as today (get_transforms()), supplemented by get_velocities() and get_accelerations() from PhysX, with COM-to-frame corrections when source/target offsets are non-identity.
Motivation
Users who drive policies, rewards, or observers from end-effector or body-relative motion (e.g. approach velocity, relative twist, acceleration limits) currently must duplicate logic: read poses from FrameTransformer, pull velocities from articulation/rigid-object APIs elsewhere, and manually align frames and offsets.
This feature would centralize pose + first/second-order kinematics for the same named frames, matching how teams already rely on FrameTransformer for relative pose.
Alternatives
- Users combine
FrameTransformer with separate Articulation / RigidObject data and implement relative velocity/acceleration themselves. More work and prone to mismatch with FrameTransformer offsets.
- Use IMU to report velocity and acceleration with offset; however, it is limited to the world frame.
Build Info
Describe the versions where you are observing the missing feature in:
- Isaac Lab Version: 2.3.2
- Isaac Sim Version: 5.1
Checklist
Acceptance Criteria
Add the criteria for which this task is considered done. If not known at issue creation time, you can add this once the issue is assigned.
Proposal
Extend
FrameTransformerso it can optionally report linear and angular velocity, and optionally linear and angular acceleration, for the source and target frames, not only pose. Data would be exposed in both world and source frames (mirroring existing*_w/*_sourcepose fields), using the same rigid-body view as today (get_transforms()), supplemented byget_velocities()andget_accelerations()from PhysX, with COM-to-frame corrections when source/target offsets are non-identity.Motivation
Users who drive policies, rewards, or observers from end-effector or body-relative motion (e.g. approach velocity, relative twist, acceleration limits) currently must duplicate logic: read poses from
FrameTransformer, pull velocities from articulation/rigid-object APIs elsewhere, and manually align frames and offsets.This feature would centralize pose + first/second-order kinematics for the same named frames, matching how teams already rely on
FrameTransformerfor relative pose.Alternatives
FrameTransformerwith separateArticulation/RigidObjectdata and implement relative velocity/acceleration themselves. More work and prone to mismatch withFrameTransformeroffsets.Build Info
Describe the versions where you are observing the missing feature in:
Checklist
Acceptance Criteria
Add the criteria for which this task is considered done. If not known at issue creation time, you can add this once the issue is assigned.
FrameTransformerData(and config) define opt-in fields for linear/angular velocity and acceleration in world and source frames, with documented shapes and conventions (including relative motion definitions).FrameTransformerdocstring) describing flags, semantics, and any PhysX/sim prerequisites for accelerations.