trajectory op• Data kinds: points × signal → table
• Call: import fullseye as fs; fs.ledger.traj_tau_dot(points, t_s, target=(0.0, 0.0, 0.0), normal=None) (to call the implementation directly, import flyvision; flyvision.traj_tau_dot(points, t_s, target=(0.0, 0.0, 0.0), normal=None); from the registry, opsflyvision.get("traj_tau_dot"))
Distance, approach speed, tau and tau-dot along a measured 3-D trajectory.
The quantity a looming detector is supposed to deliver, computed from the true path so that an eye's estimate can be scored against it. With d(t) the distance to *target* (a point, or — with *normal* — the signed distance to the plane through *target*), the approach speed is v = -dd/dt, the time-to-contact tau = d/v and its rate tau-dot = d tau/dt. Lee's (1976) sign convention: tau-dot = -1 at constant speed, tau-dot = -0.5 under the constant deceleration that stops exactly at the target (the braking law), and a value between them means the agent is braking enough to stop short.
points: (N, 3) positions, metres, N >= 3.
t_s: (N,) strictly increasing sample times, seconds.
target: the point approached (3,), metres.
normal: None for the Euclidean distance to *target*, or a (3,) plane
normal: the distance is then (p - target) . normal / |normal|.
Returns a dict {"t_s", "distance_m", "speed_m_s", "tau_s", "tau_dot", "closest_m", "t_closest_s", "n"}. Samples where the agent is not approaching (v <= 0) carry NaN in tau_s and tau_dot — a documented non-finite: a receding agent has no time-to-contact.
Ground truth (pinned in the tests): for d = v0 t toward the target at constant speed tau-dot == -1, and for d = v0 t - a t^2 / 2 with v0^2 = 2 a d0 (stops exactly at the target) tau-dot == -0.5 — both to 1e-9 on a uniform grid, where central differences are exact for a quadratic.
Raises ValueError: *points* not (N, 3) with N >= 3 (an empty *points* is refused), a *t_s* that is not (N,) strictly increasing, non-finite entries, a malformed *target* / *normal*, and a zero *normal*.
• 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_zebra_stripes_fly_eye — py -3.11 examples/poc_zebra_stripes_fly_eye.py
table as input)fly_t4t5_field · fly_flow_from_directions · fly_matched_filter · fly_egomotion_from_flow · fly_eye_merge · fly_view_plane_texture · fly_hex_resample · fly_expansion_readout
trajectory)—
*Provenance: flyvision.py — FLYVISION 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.