Ticket 086: Stochastic Propagator Module Split

Status

Implemented.

Goal

Split estimator/execution/propagator.py into focused internal modules while preserving the public run_stochastic_propagation behavior exactly.

Motivation

The stochastic propagator has absorbed particle sampling, true-state tracking, estimated-state tracking, closed-loop controller snapshots, timeline building, summary statistics, and policy reserve calculations. That makes future work such as stochastic GeoJSON export, progress feedback, and divert-route layers harder to review safely.

This is too risky for a drive-by cleanup because stochastic outputs are seed-sensitive and several tickets rely on exact reproducibility. The split should be done as a dedicated behavior-preserving refactor with broad regression coverage.

Scope

Keep estimator/execution/propagator.py as the public facade for run_stochastic_propagation, then move private implementation pieces into a small package such as estimator/execution/propagation/:

Module	Responsibility
curves.py	Energy and position interpolation curves
particles.py	Particle track and population dataclasses
sampling.py	Sampled-parameter draws and particle creation
timeline.py	Timeline advancement and snapshot construction
stats.py	Sample statistics, rates, and reserve helpers

The exact module names can differ if the implementation finds a clearer local fit, but responsibilities should stay separated.

Non-Goals

• Do not change stochastic schemas or envelope formats.
• Do not change CLI behavior, exit codes, or output rendering.
• Do not alter RNG draw order, seed handling, or floating-point operations.
• Do not rewrite the EKF or closed-loop controller logic.

Files to Create or Modify

File	Change
estimator/execution/propagator.py	Reduce to facade/orchestration and imports
estimator/execution/propagation/__init__.py	New internal package marker
estimator/execution/propagation/curves.py	New curve helpers
estimator/execution/propagation/particles.py	New particle dataclasses
estimator/execution/propagation/sampling.py	New particle sampler helpers
estimator/execution/propagation/timeline.py	New timeline builder
estimator/execution/propagation/stats.py	New statistics helpers
tests/test_stochastic_propagator.py	Add exact-output regression coverage if needed
tests/test_observation_model.py	Confirm twin-state behavior remains unchanged
docs/tickets/README.md	Mark implemented when done

Acceptance Criteria

1. run_stochastic_propagation remains importable from the same module and has the same public signature.
2. Existing stochastic, observation-model, and closed-loop-control tests pass without fixture updates.
3. A same-seed stochastic run before and after the split produces identical StochasticPropagationResult values, including timelines and policy rates.
4. When vehicle.sensors is None, the no-sensor path remains bit-for-bit compatible with the current implementation.
5. New internal modules do not import CLI or adapter code.
6. The split leaves no module with mixed sampling, timeline rendering, and stats responsibilities.