Repository navigation
Description
Task
T-076· Milestone M6 — Analysis + risk · Size S · Priority P0 · Epic #6Implements FR-RSK-001 · FR-RSK-002 · FR-RSK-003 · PR-07
Branch
feat/t-076-riskmapper-p-fail-slices-for— one task = one branch = one PR (doc 32 §6)
Summary
Implement risk/mapper.py per doc 05 §8 / doc 09 §5: p_fail = Φ(−μ/σ_tot), regions SAFE/UNCERTAIN/FAILURE from configurable thresholds, slice(x, y, anchor, resolution) on a grid with others at nominal or a mode medoid, all_pair_slices, classify.
Context / motivation
Produces the Risk Atlas — the operating-envelope map.
Scope
In scope
RiskSlicemodel incl.grid_u, physical axis values (log for illuminance),p_fail,mean,std,region..npzserialisation helper.RiskThresholdssettings model validated0 < safe < failure < 1.
Implementation guidance
- Module
roboprobe/risk/mapper.py— interface exactly doc 05 §8:class RiskMapper: def __init__(self, gp: RobustnessGP, space: ScenarioSpace, thresholds: RiskThresholds); def slice(self, x_param: str, y_param: str, anchor: Mapping[str, Any], resolution: int = 50) -> RiskSlice; def all_pair_slices(self, anchor, resolution: int = 50) -> list[RiskSlice]; def classify(self, X_unit: NDArray) -> NDArray[np.int8]. p_fail = Φ(−μ/σ_tot)withσ_tot = sqrt(σ² + σ_noise²)(doc 09 §5.1); regions: SAFE ifp_fail < thresholds.safe, FAILURE if> thresholds.failure, else UNCERTAIN.- Grid in unit space
linspace(0,1,resolution); physical axes via denormalise (log axis flag for log-scaled params); remaining parameters fixed to anchor (nominal or a mode medoid), pins respected. - Batch prediction: build all pair grids in one tensor (15 pairs × 2 500 points for d = 6) and call
gp.predictonce (torch.no_grad()). RiskSlice(frozen):x_param,y_param,x_values,y_values,log_x,log_y,p_fail,mean,std,region(int8 grid),anchor..npzserialisation helperssave_slices(path, slices)/load_slices(path)(used by T-075a and the API cache).- Standards: respect the layering contracts (doc 16 §2) and conventions in doc 34; no hard-coded tunables — thresholds/defaults come from typed settings (doc 27); no magic numbers (ruff
PLR2004); public APIs typed (mypy strict) and documented.
Configuration keys
From the configuration reference (doc 27). Keys marked (new) must be added to the typed settings models and appear in the generated reference (T-003c).
risk.thresholdsrisk.slice_resolution
Acceptance criteria
- Grid shapes and log axes correct.
- Classification thresholds honoured; pins respected.
- 15 slices for d = 6 computed in < 1 s on CPU (benchmark).
-
classifyoutput codes match doc 05 §8 (0/1/2). - npz round-trip equality test.
- Tests listed below added and green in CI; coverage of new code meets the gate in doc 18 §3.
- Docs, config reference and CHANGELOG
Unreleasedupdated where behaviour, keys or interfaces changed. - Definition of Done (doc 32 §3) satisfied.
Tests required
tests/unit/risk/test_mapper.py.- Stub GP returning analytic μ/σ to test classification exactly.
Dependencies
Blocked by: #110 (T-055a)
Blocks: #136 (T-077)
Parallelization notes
Files / paths owned by this task — avoid editing other paths; shared files (pyproject.toml entry points, cli/main.py, api/routers/__init__.py, router.tsx, docker/compose.yaml, ci.yml) are append-only:
backend/src/roboprobe/risk/mapper.py
Parallel with analysis tasks.
Definition of Done
Per doc 32 §3 and AGENTS.md §5: lint, format, mypy strict and import-linter green; tests added and passing; docs updated; PR template completed with Closes #<this issue>; Conventional Commit messages referencing T-076 (doc 16 §6); commits authored only by the repository owner identity (AGENTS.md §4); CI green; reviewer checklist (doc 32 §5.1) satisfied.
References
Design docs: doc 05 §8 · doc 09 §5.1 · doc 09 §5.2 · doc 09 §5.3 · doc 02 §3.6
Requirements: FR-RSK-001 · FR-RSK-002 · FR-RSK-003 · PR-07 (traceability: doc 28)
Metadata
Metadata
Assignees
Labels
Projects
- StatusShow more project fieldsBacklog
Activity