ball op• Data kinds: signal × points × table → table
• Call: import fullseye as fs; fs.ledger.ballistic_vs_thrust(t, p, bp: 'dict', *, iters: 'int' = 20, dt: 'float' = 0.001) -> 'dict' (to call the implementation directly, import ballistics; ballistics.ballistic_vs_thrust(t, p, bp: 'dict', *, iters: 'int' = 20, dt: 'float' = 0.001) -> 'dict'; from the registry, opsdrive.get("ballistic_vs_thrust"))
• GPU: none (CPU reference implementation only)
Fits both "gravity + drag" (drag_fit) and "gravity + drag + constant specific thrust" (thrust_fit) to the same trajectory and returns the AIC difference (aic_from_rss) ΔAIC = AIC_ballistic − AIC_thrust. Returns {"ballistic", "thrust"} (the two fits) plus "aic_ballistic", "aic_thrust", "delta_aic", "thrust_norm". No verdict is returned (the imgforensics convention): read it by parsimony — the thrust model has 3 more parameters and lowers RSS slightly even on pure noise, so unless it wins by more than the penalty 2Δk the simpler (ballistic) model is kept. Noise-only ΔAIC scatters with −2Δk as its lower bound and roughly a χ²(Δk) width (median about −2Δk + 1; strictly a mixture, since some fits hit the wall C_d = 0), and because the thrust fit starts from the ballistic solution ΔAIC ≥ −2Δk holds. The customary "decisive" value is 10 (false positives well below 1 %), but no threshold is bundled; the caller decides. When the thrust fit's RSS is below 10× the discretisation floor rss_floor (a noise-free track, or one resolved finer than the grid) it raises ValueError — AIC compares noise, not rounding (the noise-free truth gave ΔAIC = +144 at the default dt, 2026-10-23).
Empty input (t with 0 elements) raises ValueError.
• Sample-data catalog (download URLs / licences) — 2-D uses skimage.data (BSD/public domain) plus synthetic images; 3-D lists download URLs for real data sources (Stanford, PDS, …).
• Operator provenance and references — the sources of the research/methods this op family came from.
• The canonical algorithm (author, year) and its uses are named in the family usage guide above.
• poc_video_physics_consistency — py -3.11 examples/poc_video_physics_consistency.py
table as input)course_layout · course_occupancy · course_contains · world_build · world_camera · world_move · lidar_scan · rss_longitudinal_same
ball)ball_params · impact_params · flight_vacuum · flight_ode · flight_simulate · flight_state_at · magnus_lift_coefficient · drag_coefficient_sphere
*Provenance: ballistics.py — DRIVE operator registry. This per-op note is generated by tools/opdocs.py md (do not hand-edit).*
© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.