Travel Time Index (TTI): Evaluation Methods, Calculation, and Validation Using Didi Trajectory Data

The Travel Time Index (TTI) is a widely used metric for evaluating urban traffic congestion. It is defined as the ratio of actual travel time to free‑flow travel time; a larger value indicates poorer traffic conditions and is generally positively correlated with congestion. Weather anomalies and abnormal road conditions can also affect TTI.

Using massive Didi trajectory data, city traffic operation indices can be accurately calculated and visualized, providing decision support for urban managers.

Evaluation methods used domestically and abroad:

1. Based on travel time : Compare actual travel time with free‑flow travel time. TomTom uses the proportion of additional time over free flow.

2. Based on travel speed : Compare actual speed with free‑flow speed. Inrix uses the ratio of free‑flow speed to actual speed, weighted by road length.

3. Based on congestion ratio : Assign weights to each road segment and compute the proportion of congested mileage.

4. Based on traffic flow : Use the ratio of observed traffic volume to a benchmark, weighted by average daily traffic per kilometre.

Computation of the traffic index:

Speed calculation: For a link of length S observed in two consecutive time slots t1 and t2, the average speed v = 2·S / (t1 + t2).

TTI calculation: For a link in a given time slot, TTI = free‑flow speed / actual speed.

When trajectory coverage is low, links are filtered based on length and confidence of road conditions.

Free‑flow speed determination:

a) Collect 24‑hour speed samples for a link at intervals ≤5 minutes over at least one month, discarding intervals with insufficient samples.

b) Compute the arithmetic mean of all valid samples within each interval.

c) Use the average of the highest‑speed four‑hour window as the free‑flow speed for the link.

Weight calculation: In practice, the cumulative vehicle count passing a link over a 30‑day period is used as the weight, since real‑time flow data are hard to obtain.

Link set acquisition:

• Polygon data : Perform spatial queries in a PostGIS road‑network database to obtain link_id and direction, then convert to esiwei_id; topology is derived via ST_Intersects .

• Line data : Apply template‑matching based on link geometry and attributes to retrieve target link collections.

PostGIS query performance: retrieving links within a nationwide network for a specified region takes about 1.7 seconds.

System architecture: The solution consists of data ingestion, preprocessing, link extraction, metric computation, and visualization modules (illustrated in the original diagrams).

Accuracy validation:

TTI is a macro‑level indicator; absolute accuracy is less important than relative consistency. Validation cases include:

• Observing known traffic patterns in software park squares over three days (2018‑07‑09, 2019‑01‑17, 2019‑01‑18) and confirming peak times align with typical work schedules.

• Comparing Beijing and Urumqi daily curves, revealing that Urumqi’s peaks shift according to local sunrise/sunset despite using Beijing time.

• Analyzing five representative days in Beijing (including Chinese New Year, first workday after holiday, a regular week, and a day with traffic control) to illustrate the impact of holidays and events on congestion.

Feedback from traffic authorities in multiple Chinese cities (Beijing, Shanghai, Chengdu, Tianjin, Zhuhai, Weifang, Harbin, etc.) confirms the practical usefulness of the Didi traffic index.

The project is open‑source; installation, usage instructions, and FAQs are available on GitHub: github.com/didi/TrafficIndex .