Q QVDFE-CBI Repository

Code documentation · July 2026

A self-contained pipeline from corridor observations to QVDFE emissions.

This site documents the implementation, stage contracts, outputs, and reproduction commands. The manuscript itself is maintained only under QVDFE_paper/current.

Browse the package Reproduction protocol

Architecture

One package, flat source layout.

All executable modules live directly under experiments/codes/qvdfe_cbi/src. The exact CBI QC and state-transition dependencies used by the experiment are vendored there with cbi_ prefixes, so no second package or external checkout is required.

Inputs

Corridor observations

I-10 and I-405 provide measured flow, speed, and density. I-17 provides speed and TMC metadata.

Traffic model

CBI and QVDF

QC, episode detection, outlier screens, QVDF calibration, ranking, and speed reconstruction remain auditable stages.

Emission model

Benchmark and two-regime closure

The code separates reconstructed-speed MOVES error propagation from exact and cubic episode-level QVDFE emissions.

Stage contracts

Each stage leaves inspectable inputs and outputs.

  1. 0

    Load and harmonize

    Read every sensor/TMC, convert units, deduplicate metadata, and verify q = kv.

  2. 1

    Quality control

    Apply physical-range, Hampel, temporal-jump, and isolated-point repair rules.

  3. 2

    FD and episodes

    Calibrate measured S3 FDs, synthesize INRIX flow, and run both episode detectors.

  4. 3

    Robust QVDF calibration

    Screen outliers and fit duration and deepest-queue speed branches.

  5. 4

    Bottleneck ranking

    Combine episode support, severity, data quality, and calibration evidence.

  6. 5

    Speed reconstruction

    Generate five-minute profiles and corridor/detector error summaries.

  7. 6–7

    Emissions

    Run the reconstructed-speed benchmark, then exact and cubic two-regime episode models and convexity diagnostics.

Reference output

The complete latest run is published under one directory.

PathPurpose
datasets/<corridor>/stage0–stage3Input audit, QC, FD context, and detected episodes
stage2b_outliersEpisode scores, exclusions, and clean calibration set
stage3_qvdfSensor/period QVDF parameters and fit statistics
stage4_rankingBottleneck rankings and AM/PM comparisons
stage5_speed_reconstructionFive-minute profiles and reconstruction errors
stage6_emissionsObserved/reconstructed-speed operating-mode benchmark
stage7_two_regime_emissionsExact/cubic episode totals, error decomposition, and monotonicity grids
logs and reportDetailed execution record and reviewer-oriented interpretation

Reproduction

Clone, verify, and rerun.

✓ Git LFS input retrieval ✓ SHA-256 input verification ✓ Self-contained Python package ✓ Detailed and terminal logs
Terminal
git lfs install
git clone https://github.com/Mmdabb/QVDFE.git
cd QVDFE
git lfs pull
python -m pip install -r requirements-lock.txt
python -m pip install -e experiments/codes/qvdfe_cbi
python scripts/check_inputs.py
python scripts/verify_publication.py
python experiments/codes/qvdfe_cbi/run_pipeline.py