Movatterモバイル変換


[0]ホーム

URL:


Skip to content

Navigation Menu

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

Sign up

10x improvement in pose reaching accuracy#436

Discussion options

Changes in Default Behavior

  • Acceleration and Jerk in Output trajectory from motion_gen is not filtered. Previously, this was
    filtered with a sliding window to remove aliasing artifacts. To get previous behavior, set
    filter_robot_command=True inMotionGenConfig.load_from_robot_config().
  • Terminal action for motion planning is now fixed from initial seed. This improves accuracy (10x).
    To get previous behavior, settrajopt_fix_terminal_action=True and also
    trajopt_js_fix_terminal_action=True inMotionGenConfig.load_from_robot_config().
  • Introduce higher accuracy weights for IK ingradient_ik_autotune.yml. To use old file,
    passgradient_ik_file='gradient_ik.yml' inMotionGenConfig.load_from_robot_config(). Similarly
    for IKSolver, passgradient_file='gradient_ik.yml' inIKSolverConfig.load_from_robot_config().

New Features

  • Add fix terminal action in quasi-netwon solvers. This keeps the final action constant (from
    initial seed) and only optimizing for the remaining states. Improved accuracy in
    reaching targets (10x improvement for Cartesian pose targets and exact reaching for joint position
    targets).

BugFixes & Misc.

  • Fix bug (opposite sign) in gradient calculation for jerk. Trajectory optimizaiton generates
    shorter motion time trajectories.
  • Fix numerical precision issues when calculating linear interpolated seeds by copying terminal
    state to final action of trajectory after interpolation.

This discussion was created from the release10x improvement in pose reaching accuracy.
You must be logged in to vote

Replies: 0 comments

Sign up for freeto join this conversation on GitHub. Already have an account?Sign in to comment
Labels
None yet
1 participant
@balakumar-s

[8]ページ先頭

©2009-2025 Movatter.jp