How We Built a Real‑Time Road‑Testing Platform to Boost Navigation Accuracy

Background and Challenges

Road testing is not simply moving a workstation into a vehicle; it requires continuous monitoring of navigation updates, instant problem recording at each intersection, and comprehensive data collection across multiple devices, which traditionally leads to chaotic and time‑consuming processes.

Problem detection is hard: Real‑time navigation updates across several devices are difficult to capture quickly.

Incomplete information capture: Instantaneous issues lack comprehensive data, making follow‑up challenging.

Data consolidation is slow: An 8‑hour road test can require an additional 4 hours for data整理.

Solution Design

We focused the platform on three core modules: Data Collection , Data Reporting , and Control Panel , each described below.

Data Collection : Connect road‑test devices via Scrcpy, WDA, OBS, etc., to reliably capture video, images, and logs.

Data Reporting : Push feedback to Feishu groups and online documents using message cards, enabling efficient issue tracking.

Control Panel : Build front‑end panels for different test scenarios through task and UI configuration, offering a flexible user experience.

These modules provide a fast‑track foundation for capability building, which we further extend with internal tool platforms to support pre‑ and post‑test activities.

Capability Construction

3.1 Data Collection

The main targets are Android, iOS test devices and in‑vehicle cameras. By packaging screen, log, and video data at the moment of an issue, we dramatically reduce manual workload. Initially we considered a cloud‑live solution (stream, store, snapshot), but it proved costly and lacked remote control, so we sought an alternative.

Core Solution

We referenced the Huolala Cloud‑Device platform and deployed a road‑test service on the in‑vehicle computer, casting device screens to a browser for data collection. This allows both viewing and controlling the mobile screen.

Android Casting Solution

Scrcpy is an open‑source cross‑platform screen‑mirroring tool that encodes the Android screen into a video stream via ADB, sends mouse/keyboard events back, and supports remote control.

public static DesktopConnection open(int scid, boolean tunnelForward, boolean video, boolean audio, boolean control, boolean sendDummyByte) throws IOException { ... }

Access Scheme

Scrcpy provides a client for multiple platforms, but we needed a Web‑based view. We built a WebSocket service that receives Scrcpy data locally and forwards it to browsers.

const express = require('express');
const http = require('http');
const WebSocket = require('ws');
// start Scrcpy and forward
const app = express();
const server = http.createServer(app);
const wss = new WebSocket.Server({ server });

wss.on('connection', ws => {
  console.log('Connected to client');
  setInterval(() => {
    const screenData = getScreenData(); // from Scrcpy
    if (screenData) {
      ws.send(screenData);
    }
  }, 100);

  ws.on('close', () => {
    console.log('Client disconnected');
  });
});

iOS Casting Solution

We used WebDriverAgent to obtain real‑time iOS screen content via WebSocket. To handle instability, we supplemented with OBS video push, which also captures audio.

// Use OBS to sync iOS video
const OBSWebSocket = require('obs-websocket-js').default;
const obs = new OBSWebSocket();
obs.connect('ws://127.0.0.1:4455', '******')
  .then(() => console.log(`Success!`))
  .catch(error => console.log('Connection failed', error));

After integration, the system can display and record vehicle‑mounted computer screens, collect Android logs via adbkit, and quickly retrieve all device information when an issue occurs.

3.2 Data Reporting

Collected data is pushed to Feishu via message cards. Suspected product defects are submitted to the quality‑management platform with one click. This improves speed of issue flow, centralizes documentation, and standardizes defect handling.

Message cards enable rapid delivery and processing.

Automatic online document generation simplifies review.

Integration with quality‑management platform standardizes defect management.

3.3 Control Panel

The control panel provides a complete, efficient workstation for road‑test staff to configure basic information, monitor device status, and instantly record and report problems.

3.4 Video Live and Recording

Real‑time streaming and recording of the in‑vehicle computer screen allow remote assistance and post‑test review. We leveraged Feishu meetings for screen sharing and recording, enabling one‑click video recording and automatic meeting launch.

During testing, the shared video also transcribes voice prompts into text, improving issue identification.

3.5 One‑Click Order

Unlike typical navigation testing, freight scenarios require many pre‑steps (order placement, acceptance, etc.). We integrated Data Factory and LBS platforms to provide a one‑click ordering solution that includes hotspot recommendation, route comparison, and batch order creation.

Automatic hotspot recommendation based on test city.

Route similarity assessment using distance, time, and shape metrics.

Batch order creation via Data Factory for rapid workflow.

Practice Application

The platform has been in production since its inception. It supports road‑testing in over ten cities, covering more than 10,000 km and collecting 1,520 feedback items. Efficiency improvements include reducing issue recording time from minutes to seconds, cutting daily report compilation from 4 hours to 10 minutes, and tripling daily feedback volume.

4.1 Efficiency Gains

Issue recording time improved from minutes to seconds.

Daily report preparation time saved 96% (4 h → 10 min).

Feedback volume increased threefold (5.5 → 17.25 per day).

Standardized, richer feedback improves test quality.

4.2 Deployment Scale

In the past year the platform supported road‑testing in more than ten cities, accumulating over 10,000 km of driving and 1,520 feedback entries, enabling teams to focus on product impact rather than manual data handling.

Future Outlook

We will continue to explore more stable and convenient service models, deepen data value extraction, and pursue breakthroughs such as interaction flow optimization, real‑time tracking of test vehicles, and comprehensive dashboards for decision support.

Interaction flow optimization: Refine processes for stability and ease of use.

Road‑test tracking: Summarize vehicle status in real time and provide remote entry points.

Platform dashboard: Full‑view statistics and metrics to guide product planning.