How Gaode Maps Calculates Traffic Light Countdown: Inside the Algorithm and Patent

Internal Employee Account

A former staff member described the daily workflow for the traffic‑light data team: each morning the team manually checks the timing of traffic signals in its assigned area, records the values by hand, and uploads the results to the backend system. This process is entirely manual, which introduces potential human error.

Algorithmic Basis

The production system relies on roughly 99% algorithmic inference and only about 1% real‑world traffic‑control data obtained through partnerships with traffic authorities. The algorithms predict the current signal phase (red, yellow, green) and also generate recommended vehicle speeds to achieve a “green‑wave” effect, thereby improving travel convenience and safety.

Patent Reference

The method for extracting traffic‑light cycle durations is documented in Chinese patent CN114463969A filed by Gaode Software Co. The patent outlines a procedure to calculate the length of a signal cycle from observed data and to derive countdown timers for each intersection.

Technical Implementation of Countdown

Implementing a countdown timer requires determining the current offset within a traffic‑light cycle. A practical approach consists of the following steps:

Use the patented algorithm to estimate the total cycle length (e.g., 90 s, 120 s) based on historical vehicle trajectories and signal‑phase inference.

Collect real‑time vehicle start‑stop information from navigation sessions (e.g., timestamps when a vehicle stops at a red light and resumes moving).

Align the vehicle’s stop‑resume timestamps with the estimated cycle to infer the phase start time.

Calculate the remaining time as remaining = cycle_length – elapsed_since_cycle_start. Optionally apply smoothing or Kalman filtering to reduce noise.

Update the countdown continuously as new vehicle telemetry arrives, ensuring the displayed time stays synchronized with the inferred signal phase.

This method allows the system to provide per‑intersection countdowns and estimated waiting‑lane counts without direct access to the traffic‑control hardware, relying instead on algorithmic inference combined with live navigation data.