traffic op• Data kinds: any → table
• Call: import fullseye as fs; fs.ledger.importance_risk_estimate(simulate_fn: 'Callable', base_rate: 'Callable', boosted_rate: 'Callable', n_runs: 'int', seed, *, x_max: 'float', boosted_rate_max: 'float', n_grid: 'int' = 20001) -> 'Dict[str, float]' (to call the implementation directly, import drivetraffic; drivetraffic.importance_risk_estimate(simulate_fn: 'Callable', base_rate: 'Callable', boosted_rate: 'Callable', n_runs: 'int', seed, *, x_max: 'float', boosted_rate_max: 'float', n_grid: 'int' = 20001) -> 'Dict[str, float]'; from the registry, opsdrive.get("importance_risk_estimate"))
Makes accidents more likely with trials at a raised emergence rate (rate λ') and maps back to the accident probability at the original rate λ with likelihood-ratio weights.
One trial = draw occurrence positions x_1..x_N with poisson_events(boosted_rate); simulate_fn(events, rng) gives 1 for a crash (a bool or a value in [0, 1]). Weight (Radon–Nikodym derivative of the Poisson process dP_λ/dP_λ'):
w = Π_i λ(x_i)/λ'(x_i) · exp(−(Λ − Λ')), Λ = ∫λ, Λ' = ∫λ' (trapezoid, n_grid points)
Estimate p̂ = mean(w · Y), standard error = std(w · Y)/sqrt(n). Setting boosted_rate equal to base_rate gives w ≡ 1, i.e. naive Monte Carlo (the gate compares against this and against a simulate_fn whose crash probability is known in closed form). If there is a place where λ > 0 but λ' = 0, the estimate is biased (not absolutely continuous), so this is checked on the grid and raises ValueError.
Return dict: estimate, std_error, per_run_var (sample variance of w·Y), n_runs, Lambda, Lambda_boost, mean_weight (should be ≈ 1; sanity of the weights), ess (effective sample size (Σw)²/Σw²), hits (number of times Y > 0).
• 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_driving_traffic — py -3.11 examples/poc_driving_traffic.py
table as input)course_layout · course_occupancy · course_contains · world_build · world_camera · world_move · lidar_scan · rss_longitudinal_same
traffic)idm_accel · idm_equilibrium_gap · idm_platoon_simulate · driver_style · lateral_wobble · ou_estimate · social_force_step · pedestrian_crossing
*Provenance: drivetraffic.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.