Skip to content

PPP-AR Experiments

Generated: 2026-04-02 21:34 UTC

This page is the current externalized state of the PPP-AR experiment lane. All rows below use the same strategy catalog, the same metrics, and comparable CLAS inputs.

Shared Interface

  • Suite: CLAS PPP phase-bias bank suite
  • Cases: 2
  • Primary mode: static
  • Shared strategy catalog: experiments/ppp_ar/strategies.toml
  • Shared output schema: wall_time_s, ppp_solution_rate_pct, ppp_fixed_solutions, fallback_solutions, clas_hybrid_fallback_epochs, mean_3d_error_m, median_3d_error_m, p95_3d_error_m, max_3d_error_m, readability_score, extensibility_score

Suite Summary

Strategy Cases passed Mean solution rate Mean 3D (m) P95 3D (m) Mean delta vs baseline (m) Fixed total Hybrid fallback total Promotion
IFLC Float Baseline 0/2 99.58 11.957 9.930 0.000 0 0 stable_control
OSR Float Strict Pipeline 0/2 50.00 4.898 5.134 7.059 0 0 exploratory
OSR Float Base-Plus-Network Phase-Bias-Composition Pipeline 0/2 50.00 4.898 5.134 7.059 0 0 exploratory
OSR Float Same-30s Phase-Bias-Bank Pipeline 0/2 50.00 4.898 5.134 7.059 0 0 exploratory
OSR Float Close-30s Phase-Bias-Bank Pipeline 0/2 50.00 4.898 5.134 7.059 0 0 exploratory
OSR Float Latest-Preceding Phase-Bias-Bank Pipeline 0/2 50.00 4.898 5.134 7.059 0 0 exploratory
OSR AR Strict Pipeline 0/2 50.00 4.898 5.134 7.059 0 0 exploratory
OSR AR Base-Plus-Network Phase-Bias-Composition Pipeline 0/2 50.00 4.898 5.134 7.059 0 0 exploratory
OSR AR Same-30s Phase-Bias-Bank Pipeline 0/2 50.00 4.898 5.134 7.059 0 0 exploratory
OSR AR Close-30s Phase-Bias-Bank Pipeline 0/2 50.00 4.898 5.134 7.059 0 0 exploratory
OSR AR Latest-Preceding Phase-Bias-Bank Pipeline 0/2 50.00 4.898 5.134 7.059 0 0 exploratory

Case Results

CLAS 2019-08-27 static PPP

  • Case key: clas_static_2019_239
  • Observation file: 0627239Q.obs
  • Navigation file: sept_2019239.nav
  • QZSS L6 source: 2019239Q.l6
  • Max epochs: 120
  • Mode: static

Strategy Comparison

Strategy Style Solution rate Fixed Fallback Hybrid fallback Mean 3D (m) P95 3D (m) Wall (s) Readability Extensibility Promotion
IFLC Float Baseline monolithic 99.17 0 1 0 4.540 4.638 5.40 low low stable_control
OSR Float Strict Pipeline pipeline+boundary 100.00 0 0 0 4.921 5.193 4.70 low low exploratory
OSR Float Base-Plus-Network Phase-Bias-Composition Pipeline pipeline+materialization 100.00 0 0 0 4.921 5.193 4.75 low low exploratory
OSR Float Same-30s Phase-Bias-Bank Pipeline pipeline+materialization 100.00 0 0 0 4.921 5.193 4.71 low low exploratory
OSR Float Close-30s Phase-Bias-Bank Pipeline pipeline+materialization 100.00 0 0 0 4.921 5.193 4.70 low low exploratory
OSR Float Latest-Preceding Phase-Bias-Bank Pipeline pipeline+materialization 100.00 0 0 0 4.921 5.193 4.70 low low exploratory
OSR AR Strict Pipeline pipeline+solver+boundary 100.00 0 0 0 4.921 5.193 4.80 low low exploratory
OSR AR Base-Plus-Network Phase-Bias-Composition Pipeline pipeline+solver+materialization 100.00 0 0 0 4.921 5.193 4.81 low medium exploratory
OSR AR Same-30s Phase-Bias-Bank Pipeline pipeline+solver+materialization 100.00 0 0 0 4.921 5.193 4.83 low medium exploratory
OSR AR Close-30s Phase-Bias-Bank Pipeline pipeline+solver+materialization 100.00 0 0 0 4.921 5.193 4.80 low medium exploratory
OSR AR Latest-Preceding Phase-Bias-Bank Pipeline pipeline+solver+materialization 100.00 0 0 0 4.921 5.193 4.83 low medium exploratory

Promotion Gate

Strategy Mean delta vs baseline (m) P95 delta vs baseline (m) Fixed delta Solution-rate delta Gate Rationale
IFLC Float Baseline 0.000 0.000 0 0.00 hold Stable control arm used to judge candidate strategies.
OSR Float Strict Pipeline -0.381 -0.555 0 0.83 hold Still exploratory because these gates are not met: mean_error_improved, p95_error_improved, readability_non_regression
OSR Float Base-Plus-Network Phase-Bias-Composition Pipeline -0.381 -0.555 0 0.83 hold Still exploratory because these gates are not met: mean_error_improved, p95_error_improved, readability_non_regression
OSR Float Same-30s Phase-Bias-Bank Pipeline -0.381 -0.555 0 0.83 hold Still exploratory because these gates are not met: mean_error_improved, p95_error_improved, readability_non_regression
OSR Float Close-30s Phase-Bias-Bank Pipeline -0.381 -0.555 0 0.83 hold Still exploratory because these gates are not met: mean_error_improved, p95_error_improved, readability_non_regression
OSR Float Latest-Preceding Phase-Bias-Bank Pipeline -0.381 -0.555 0 0.83 hold Still exploratory because these gates are not met: mean_error_improved, p95_error_improved, readability_non_regression
OSR AR Strict Pipeline -0.381 -0.555 0 0.83 hold Still exploratory because these gates are not met: mean_error_improved, p95_error_improved, readability_non_regression, fixed_epochs_improved
OSR AR Base-Plus-Network Phase-Bias-Composition Pipeline -0.381 -0.555 0 0.83 hold Still exploratory because these gates are not met: mean_error_improved, p95_error_improved, readability_non_regression, fixed_epochs_improved
OSR AR Same-30s Phase-Bias-Bank Pipeline -0.381 -0.555 0 0.83 hold Still exploratory because these gates are not met: mean_error_improved, p95_error_improved, readability_non_regression, fixed_epochs_improved
OSR AR Close-30s Phase-Bias-Bank Pipeline -0.381 -0.555 0 0.83 hold Still exploratory because these gates are not met: mean_error_improved, p95_error_improved, readability_non_regression, fixed_epochs_improved
OSR AR Latest-Preceding Phase-Bias-Bank Pipeline -0.381 -0.555 0 0.83 hold Still exploratory because these gates are not met: mean_error_improved, p95_error_improved, readability_non_regression, fixed_epochs_improved

CLAS 2018-11-24 static PPP

  • Case key: clas_static_2018_329
  • Observation file: 0161329A.obs
  • Navigation file: tskc2018329.nav
  • QZSS L6 source: 2018328X_329A.l6
  • Max epochs: 120
  • Mode: static

Strategy Comparison

Strategy Style Solution rate Fixed Fallback Hybrid fallback Mean 3D (m) P95 3D (m) Wall (s) Readability Extensibility Promotion
IFLC Float Baseline monolithic 100.00 0 0 0 19.374 15.223 14.68 low low stable_control
OSR Float Strict Pipeline pipeline+boundary 0.00 0 120 0 4.875 5.076 14.65 low low exploratory
OSR Float Base-Plus-Network Phase-Bias-Composition Pipeline pipeline+materialization 0.00 0 120 0 4.875 5.076 14.43 low low exploratory
OSR Float Same-30s Phase-Bias-Bank Pipeline pipeline+materialization 0.00 0 120 0 4.875 5.076 14.90 low low exploratory
OSR Float Close-30s Phase-Bias-Bank Pipeline pipeline+materialization 0.00 0 120 0 4.875 5.076 14.54 low low exploratory
OSR Float Latest-Preceding Phase-Bias-Bank Pipeline pipeline+materialization 0.00 0 120 0 4.875 5.076 14.46 low low exploratory
OSR AR Strict Pipeline pipeline+solver+boundary 0.00 0 120 0 4.875 5.076 14.51 low low exploratory
OSR AR Base-Plus-Network Phase-Bias-Composition Pipeline pipeline+solver+materialization 0.00 0 120 0 4.875 5.076 14.45 low medium exploratory
OSR AR Same-30s Phase-Bias-Bank Pipeline pipeline+solver+materialization 0.00 0 120 0 4.875 5.076 14.49 low medium exploratory
OSR AR Close-30s Phase-Bias-Bank Pipeline pipeline+solver+materialization 0.00 0 120 0 4.875 5.076 14.24 low medium exploratory
OSR AR Latest-Preceding Phase-Bias-Bank Pipeline pipeline+solver+materialization 0.00 0 120 0 4.875 5.076 14.25 low medium exploratory

Promotion Gate

Strategy Mean delta vs baseline (m) P95 delta vs baseline (m) Fixed delta Solution-rate delta Gate Rationale
IFLC Float Baseline 0.000 0.000 0 0.00 hold Stable control arm used to judge candidate strategies.
OSR Float Strict Pipeline 14.498 10.147 0 -100.00 hold Still exploratory because these gates are not met: solution_rate_non_regression, readability_non_regression
OSR Float Base-Plus-Network Phase-Bias-Composition Pipeline 14.498 10.147 0 -100.00 hold Still exploratory because these gates are not met: solution_rate_non_regression, readability_non_regression
OSR Float Same-30s Phase-Bias-Bank Pipeline 14.498 10.147 0 -100.00 hold Still exploratory because these gates are not met: solution_rate_non_regression, readability_non_regression
OSR Float Close-30s Phase-Bias-Bank Pipeline 14.498 10.147 0 -100.00 hold Still exploratory because these gates are not met: solution_rate_non_regression, readability_non_regression
OSR Float Latest-Preceding Phase-Bias-Bank Pipeline 14.498 10.147 0 -100.00 hold Still exploratory because these gates are not met: solution_rate_non_regression, readability_non_regression
OSR AR Strict Pipeline 14.498 10.147 0 -100.00 hold Still exploratory because these gates are not met: solution_rate_non_regression, readability_non_regression, fixed_epochs_improved
OSR AR Base-Plus-Network Phase-Bias-Composition Pipeline 14.498 10.147 0 -100.00 hold Still exploratory because these gates are not met: solution_rate_non_regression, readability_non_regression, fixed_epochs_improved
OSR AR Same-30s Phase-Bias-Bank Pipeline 14.498 10.147 0 -100.00 hold Still exploratory because these gates are not met: solution_rate_non_regression, readability_non_regression, fixed_epochs_improved
OSR AR Close-30s Phase-Bias-Bank Pipeline 14.498 10.147 0 -100.00 hold Still exploratory because these gates are not met: solution_rate_non_regression, readability_non_regression, fixed_epochs_improved
OSR AR Latest-Preceding Phase-Bias-Bank Pipeline 14.498 10.147 0 -100.00 hold Still exploratory because these gates are not met: solution_rate_non_regression, readability_non_regression, fixed_epochs_improved

Structural Heuristics

Strategy Files LOC Max file LOC Args CLASLIB reference
IFLC Float Baseline 1 4097 4097 0 Control arm used to measure how far CLASLIB-inspired observation-space work actually moves the solver.
OSR Float Strict Pipeline 3 6145 4097 5 Uses the CLASLIB-style observation-space boundary directly, without falling back to the standard PPP epoch path.
OSR Float Base-Plus-Network Phase-Bias-Composition Pipeline 4 8871 4097 5 Used to test whether the compact-to-expanded SSR lane should follow CLASLIB-style add_base_value_to_pbias semantics instead of using raw subtype-6 network values directly.
OSR Float Same-30s Phase-Bias-Bank Pipeline 4 8871 4097 5 Used to test whether the compact-to-expanded SSR lane should follow the CLASLIB-style 30-second base-bank anchoring implied by add_base_value_to_pbias instead of only using the current pending epoch.
OSR Float Close-30s Phase-Bias-Bank Pipeline 4 8871 4097 5 Used to test whether the compact-to-expanded SSR lane should follow CLASLIB-style close-bank lookup semantics rather than requiring the exact pending anchor or accepting any older preceding bank.
OSR Float Latest-Preceding Phase-Bias-Bank Pipeline 4 8871 4097 5 Used as a negative control to test whether ambiguity quality changes simply because the compact-to-expanded SSR lane reuses the most recent base phase-bias bank rather than the exact 30-second anchor.
OSR AR Strict Pipeline 4 6965 4097 8 Measures the CLASLIB-style observation-space boundary directly so hybrid fallback arms can be judged against a strict CLAS control.
OSR AR Base-Plus-Network Phase-Bias-Composition Pipeline 5 9691 4097 6 Used to test whether ambiguity fixing recovers once the compact-to-expanded SSR lane follows CLASLIB-style add_base_value_to_pbias semantics for network phase-bias rows.
OSR AR Same-30s Phase-Bias-Bank Pipeline 5 9691 4097 6 Used to test whether ambiguity fixing recovers once the compact-to-expanded SSR lane follows the CLASLIB-style 30-second base-bank anchoring implied by add_base_value_to_pbias.
OSR AR Close-30s Phase-Bias-Bank Pipeline 5 9691 4097 6 Used to test whether ambiguity fixing recovers once the compact-to-expanded SSR lane follows CLASLIB-style close-bank lookup semantics rather than requiring the exact pending anchor or accepting any older preceding bank.
OSR AR Latest-Preceding Phase-Bias-Bank Pipeline 5 9691 4097 6 Used as a negative control to test whether ambiguity fixing changes simply because the compact-to-expanded SSR lane reuses the most recent base phase-bias bank rather than the exact 30-second anchor.

Notes

  • IFLC Float Baseline is the control arm and remains the fallback path.
  • OSR * Pipeline arms keep the CLASLIB-inspired observation-space boundary and differ in how epoch fallback, atmosphere-token selection, and ambiguity fixing are staged.
  • Strict arms stay inside the CLAS epoch path. Hybrid arms restore the pre-epoch state and re-run the stable standard PPP path when the CLAS OSR stage cannot complete.
  • Guarded arms keep the nearest grid but reject stale atmosphere tokens completely once they age out.
  • Balanced arms are stale-aware: they keep the nearest grid while corrections are fresh, then prefer freshness once the current network ages out.
  • Freshness arms prioritize correction recency over grid proximity so selector failures can be isolated from the cssr2osr-style measurement path.
  • Clock-Bound-* arms keep the same accepted CLAS update path but tighten expanded-SSR timing acceptance relative to the selected clock epoch.
  • Orbit-* Source arms change the compact-to-expanded row emission policy before the solver sees any CLAS SSR values.
  • *Atmos-Merge* arms change how compact STEC polynomial terms are carried or reset before expanded SSR rows are materialized.
  • *Subtype-Merge* arms change how subtype 8/9/12 atmosphere families replace or coexist before expanded SSR rows are materialized.
  • *Phase-Bias-Merge* arms change how compact subtype-5/6 phase-bias rows reset or prune stale satellite scope before expanded SSR rows are materialized.
  • *Phase-Bias-Source* arms keep the same compact epoch contents but change whether subtype-5, subtype-6, or raw arrival order wins when both provide the same signal row before expanded SSR materialization.
  • *Phase-Bias-Composition* arms keep the same accepted source rows but change whether subtype-6 network rows are used directly, added on top of base rows, or replaced by base rows before expanded SSR materialization.
  • *Phase-Bias-Bank* arms keep the same accepted source rows and composition rule but change whether subtype-6 network rows can look up base phase-bias banks only inside the pending epoch, at the same 30-second anchor, from the closest preceding anchor within 30 seconds, or from the latest preceding anchor before expanded SSR materialization.
  • *Value* arms change how expanded atmosphere values are constructed from polynomial terms and residual lists before residual formation.
  • *Residual-Sampling* arms keep the same composed atmosphere model but change whether expanded residual lists use indexed samples, mean fallback, or mean-only sampling.
  • *Subtype12-Value* arms keep residual sampling fixed and only change how subtype-12 rows retain constant, linear, or higher-order surface terms before residual formation.
  • *Phase-Bias-Only* and *Compensation-Only* arms keep continuity timing fixed and only change whether accepted CLAS updates use compact phase-bias values, derived phase compensation, or both before SIS/repair injection.
  • *Reference-Time* arms keep accepted phase-bias values fixed and only change which epoch anchors SIS continuity and repair-offset state before phase-bias corrections are injected.
  • Promotion is decided against the control arm and only becomes stable after repeated suite wins.
  • Readability and extensibility are heuristic scores generated from experiment-owned files, file size, and command surface. They are not code-review substitutes.