I-24 MOTION RESEARCH

Wave Formation Observatory

Microscopic attribution of stop-and-go-wave formation

Research units: km · km/h · s · acceleration m/s². Road coordinates retain the original dataset origin, converted to kilometres.

ANALYSIS PIPELINE · PART 01

Detect the waves before tracing their origins.

PRIMARY DETECTOR PARAMETER

Critical speed threshold

24.14 km/h 6.7 m/s
16.09 km/hPaper's stable range32.19 km/h

Published November 22 results are loaded at the authors' 24.14 km/h baseline.

Components—current lane/filter
Median duration—seconds
Median front span—km
Complex topology—share of components
Advanced detector settings

Shown for reproducibility. These structural parameters should be varied in a later robustness study, after the critical-speed analysis.

Detected wave components

Download all dates & onset windows (CSV)

Ji et al. graph method · vc = 24.14 km/h · Lane 1

Component identityWave tailOnset seed
Click a wave component to inspect itTime after 06:00 CST

Traffic speed heatmap

Loading speed field…

0 km/h136.8+ km/h
Loading speed field…
Gray: insufficient coverageHover for speed · selected wave and onset window linked above

02 · ONSET AREA

Define the candidate search region

Explore earlier waves in broad context, then inspect physical interactions in a focused window. Each area has independent ranges around the selected onset seed.

Waiting for selectionSelect a wave in Part 01

BROAD CONTEXT

Trace the surrounding traffic history

Explore earlier waves and neighboring lanes. The amber rectangle marks the focused interaction area.

Traffic context heatmap

Select a located wave in 01.

Speed field · 8 s × 0.03219 km
Red: slow · Green: fast · Gray: insufficient coverage or outside observationDark: wave fronts · Amber: focused area in the selected onset lane

COMPUTED SCREENING · PROVISIONAL RULES

Analyze earlier-wave connections

Measure prior fronts, extrapolated paths, intervening speeds, and coverage in the chosen broad area.

Adjust screening thresholds

Prior fronts: at least 3 points spanning 24 s; fit at most the last 120 s of observations; upstream wave speed +8.05 to +40.23 km/h in kilometre coordinates; fit RMSE ≤0.06437 km; extrapolation ≤180 s. Path coverage ≥90%. Continuation-compatible: ≥80% of valid path samples ≤ the slow-speed threshold, without the specified recovery run. History check: up to 180 s at the onset position. Sustained congestion: ≥90% slow samples without recovery. Cross-lane geometry alone never establishes transfer. These are transparent initial heuristics, not validated causal thresholds.

Select a wave and run the screening. Results will include measurements and a next step.

FOCUSED INTERACTION

Inspect vehicle behavior and interactions

Adjust this window independently, then load the physical trajectories.

Physical trajectory inspection

Choose a detected component and open its investigation window.

Original data awaiting connection
Target stream · adjacent lanes included
Lane 1Decreasing kilometre coordinateJi detector direction
Linked variable views
Lane 1 · blue / Lane 2 · teal / Lane 3 · violet / Lane 4 · orange
No physical trajectories loadedThe interface is ready; reconnect the I-24 MOTION drive to extract vehicles intersecting this space–time window.
Position over time · selected window shown in amberCandidate vehicles —

Selected vehicle

Load the physical trajectories, then click a track or choose a vehicle.

Quality of the focused area

Not measured. Load the focused-area trajectories.
Inspect this quality report in 03 →

COMPUTED EVENT SEQUENCES · PILOT RULES

Wave-generation event codes 0–8 · rules and results

Read each coding rule, its measured evidence and result here. An initial ego candidate near the onset is selected automatically when trajectories load; you can change it below.

The reference vehicle is an investigation candidate, not an identified cause.

Sequence tree · corrected vehicle relationships

Event vocabulary · codes 0–8

Advanced processing settings

Local lane geometry

Fixed lateral bins are not validated lane boundaries. Confirm local calibration before enabling lane-entry labels. The default coordinates reproduce the existing display bins.

Load trajectories and select a reference vehicle.

Results will show event codes, measurements, thresholds, candidate links and missing evidence. No cause has been assigned.

Inspect the quality of this evidence

Separate source gaps, track fragmentation, and missing interaction history.

Continue to 03 Data preprocessing →

Interpretation: this rectangle is an investigation window, not a causal boundary. Vehicles that intersect it become candidates; attribution requires their preceding trajectories and behavioural evidence. Tracks are separated at detected native-sample gaps or invalid observations. Derived speed and acceleration are exploratory and are not computed across those breaks. Leader–follower links are not yet validated.

03 · DATA PREPROCESSING

Inspect the evidence before interpreting it.

Audit coverage, diagnose missing observations, and validate repairs before mechanism analysis.

← Adjust onset window
01 · Source auditInitial sample completed
02 · Gap diagnosisNot started · interaction level
03 · ReconstructionNot started
04 · ValidationNot started
05 · Evidence assessmentNot started
Download sample audit ↓

Loading source audit…

Source sampling, missing vehicle tracks, and missing interaction evidence are different questions. Regular timestamps do not demonstrate complete coverage. Event analysis reconstructs eligible internal gaps only when run; estimates and unresolved evidence are reported separately below Focused interaction.

01

Detect waveUse Ji et al.'s published 24.14 km/h baseline.

→
02

Select onsetChoose one component and its earliest seed.

→
03

Set search regionAdjust time and position asymmetrically.

→
04

Match vehiclesLoad physical trajectories intersecting the window.