tactorque opimage2d × image2d × scalar × scalar × scalar × scalar × scalar × scalar → tableimport fullseye as fs; fs.ledger.grasp_torque_frame(m_ref, m_cur, dark: 'float', pitch_px: 'float', r_px: 'float', pitch: 'float', G: 'float', nu: 'float', a: 'float | None' = None, window=None, coef: 'float | None' = None, P: 'float | None' = None, mu: 'float | None' = None, torsion_model: 'str' = 'partial_slip') -> 'dict' (実装を直接呼ぶなら import tactorque; tactorque.grasp_torque_frame(m_ref, m_cur, dark: 'float', pitch_px: 'float', r_px: 'float', pitch: 'float', G: 'float', nu: 'float', a: 'float | None' = None, window=None, coef: 'float | None' = None, P: 'float | None' = None, mu: 'float | None' = None, torsion_model: 'str' = 'partial_slip') -> 'dict'、台帳から引くなら opsdrive.get("grasp_torque_frame"))像から 1 回で: 基準・現在のマーカー像(:func:tacslip.marker_image)→ :func:tacslip.marker_track → m 単位の場 → :func:torque_decompose。
pitch_px・r_px = マーカー格子と半径 [px]、pitch = m/px、window は [px] の (cx, cy, R)。マーカー 1 個あたりの面積 = (pitch_px·pitch)²、
発散の近傍半径 = 1.5 ピッチ。P・mu・torsion_model はねじりの換算(:func:torque_decompose、既定は Hertz 接触の部分滑りで
P と mu が要る、平頭押し込み子は "no_slip")。返り = torque_decompose の表 + track(matched・n0・n1・p0・u [px])。
Raises ValueError: 追跡の対応が 9 個未満(双極子に足りない)、pitch ≤ 0、torque_decompose の ValueError。
py -3.11 examples/poc_tactile_dipole_torque.pytable を入力に取れる)course_layout · course_occupancy · course_contains · world_build · world_camera · world_move · lidar_scan · rss_longitudinal_same
tactorque)punch_pressure · punch_surface_uz · hertz_pressure_shifted · ellipse_pressure_shifted · pressure_first_moment · boussinesq_kernel · boussinesq_surface_displacement · surface_divergence_closed_form
Provenance: tactorque.py — DRIVE operator registry. この per-op ノートは tools/opdocs.py md が自動生成(手編集しない)。
© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.