[Afloat16]

Problem

Particle resampling chooses the particle whose cumulative weight is nearest to a uniform draw. That does not sample the intended categorical distribution: for weights [0.8, 0.2], the boundary is 0.9 rather than 0.8. Repeated CDF entries can also select zero-weight particles. This biases the next particle generation.

Change

Use the inverse CDF intervals via searchsorted(..., side="right"). Normalising by the last cumulative sum makes the final CDF entry exactly one. Remove the interpolation dependency that this code no longer needs.

This corrects resampling probabilities and can change seeded particle-filter trajectories.

The regressions use a seeded generator, two independently specified probability thresholds, and intervening zero-weight particles. Expected particle indices are computed directly from those thresholds.

Validation

• Both new regression cases fail on the original implementation.
• PYTHONPATH=src python -m pytest tests/test_particle_filter_api.py tests/test_particle_heading_moments.py -q: 41 passed, with 3 animation cleanup warnings.
• PYTHONPATH=src python -m pytest tests/test_vehicle_estimators.py -k TestParticleFilter -q: 7 passed, 40 deselected, checking the corrected resampling against the existing estimator tracking regressions.
• Changed-file formatting and git diff --check passed.
• Ruff reports the same three existing wildcard-import diagnostics in the module's demonstration block as on the base revision; the regression file passes Ruff.

[codacy-production]

Up to standards ✅

🟢 Issues 0 issues

Results:
0 new issues

View in Codacy

🟢 Metrics 0 complexity · 0 duplication

Metric	Results
Complexity	0
Duplication	0

View in Codacy

_{NEW Get contextual insights on your PRs based on Codacy's metrics, along with PR and Jira context, without leaving GitHub. Enable AI reviewer
TIP This summary will be updated as you push new changes.}