Trajectory Analysis
We analyzed 1,262,687 records from 6,805 vehicles. Average number of points per vehicle is 185, maximum 5,909.
Examples:
Car 1: 316 points, average speed 13.9 km/h, distance 342.7 km
Car 2: 263 points, average speed 7.56 km/h, distance 161.5 km
Map Segmentation
The map is divided into cells of 10-20 meters. For each cell we store: ID, center coordinates, latitude and longitude boundaries. This allows us to analyze movement with precision down to specific road sections.
Congestion Calculation Algorithm
We use the 90th percentile of speeds as base speed:base_speed = np.percentile(speeds[speeds > 0], 90)
Then calculate congestion coefficient:congestion = (base_speed - avg_speed) / base_speed
Result: number from 0 (free flow) to 1 (almost stopped).
Handling Special Cases
If base speed < 15 km/h (courtyards, parking lots) → set minimum congestion 0.1
If points < 5 → cell is excluded from analysis
This ensures data quality and meaningful results.
Additional Metrics
Speed variation coefficient = movement stability
Point density = statistical reliability
More data = more accurate prediction.
These metrics help validate our congestion analysis.
Visualization and Results
We build an interactive map:
■ Green — free flow (0.0–0.3)
■ Yellow — moderate congestion (0.3–0.6)
■ Red — heavy congestion (0.6–1.0)
Also includes speed map — from blue (slow) to red (fast).