Baidu AI Cloud Public Sector China Translated from Chinese

China Mars Exploration Project (中国火星探测工程)

China's Mars exploration program wanted a more immersive way to communicate with the public and explain the Zhurong rover's mission, but needed lifelike interaction, expressive animation, and multilingual conversational ability to make a science-outreach digital character convincing.

Real-time expression and lip-movement prediction accuracy exceeded 98.5%.

Reported impact

Business outcome

The Mars exploration program gained a reusable digital-human outreach channel that can explain mission details interactively, make science communication more engaging, and broaden public interest in aerospace exploration.

  • Customer Experience
  • Speed to Market

Solution

Baidu AI Cloud and the China Mars Exploration Project created a digital-human version of the Zhurong rover to turn aerospace outreach into a live interactive experience. The solution combines 4D scanning, high-fidelity 3D character production, lightweight neural-network driving, and multimodal AI for speech and language. In practice, the digital human can move, blush, flap its 'wings,' describe its instruments, and respond to audience questions during public events. Under the hood, Baidu's system binds scanned training data to 3D models, then uses machine learning to drive expressions and lip movements in real time with low latency. The same stack also mixes prebuilt expressions with AI-generated animation and supports multiple input modes, including text, voice, RGB camera facial control, and depth-camera capture. That gives the Mars program a reusable digital-human channel for science communication that feels more like conversation than scripted video playback.

Data flow

Visual and motion data captured through 4D scanning is bound to the Zhurong digital-human model, where neural-network driving predicts expression and lip movement. Text, speech, or camera-driven input then triggers the platform's multimodal language stack, and the resulting animation and responses are rendered back to live audiences in real time.

Solution architecture

4 components · 3 layers
  1. Ingest
    • 4D Scanning Technology Captures the high-precision training data needed to build and tune the digital human's visual behavior.
  2. Compute
    • Multimodal AI Speech and Language Stack Supports multilingual speech, understanding, and interaction for the digital-human outreach experience.
  3. Serving
    • 3D Avatar Production Technology Creates the rover's lifelike 3D body, expressions, and motion foundation for interactive presentation.
    • Xiling Intelligent Digital Human Platform Runs the real-time digital-human experience and coordinates animation, dialogue, and rendering.

Architecture clues

  • Multiple driving modes are supported, including text, speech, RGB-camera control, and depth-camera capture.
  • The digital human is positioned as a reusable science-communication channel, not just a one-off event animation.
  • The solution blends prebuilt expressions with AI-generated expressions in real time rather than relying on canned animation only.

Evidence from the source

A lightweight neural network predicts expressions and lip movements in real time with accuracy above 98.5% (original: 采用了轻量级的深度神经网络模型,能够实现端到端的表情和口型实时预测,准确率高于98.5%).
Baidu and the China Mars Exploration Project jointly unveiled the world's first digital human for a Mars rover (original: 中国火星探测工程联合百度发布的,全球首个火星车数字人).
The digital human uses 4D scanning and high-fidelity 3D character-production technology (original: 百度智能云采用国内首创的4D扫描技术…基于百度智能云的影视级高精拟真3D人像制作技术、拟人化3D卡通形象制作技术).
The system supports text-driven, voice-driven, RGB-camera facial driving, and depth-camera driving modes (original: 百度的数字人技术拥有文本驱动、语音驱动、普通 RGB 摄像头面部驱动、深度摄像头面部采集驱动共四种驱动方式).
Transparency Evidence & methodology 86/100 reference value · 89% confidence · 1 source

Review how this reference was scored and trace each normalized claim back to its documented source.

01

Scoring methodology

How scores are calculated

Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.

Customer reference value86/100
  • Problem clarity20/20
    Measures whether the customer's operational problem is specific enough to compare.
  • Implementation depth25/25
    Rewards named components, data flow, integrations, and architecture detail.
  • Outcome specificity16/20
    Measures whether the reported benefit is concrete rather than a generic claim.
  • Quantified impact15/25
    1 metric describe measurable impact.
  • Deployment scale10/10
    Rewards evidence of a live or scaled deployment and meaningful integration scope.
Evidence confidence89%
  • Source authority20/25
    The main evidence is an official vendor-published customer story.
  • Claim traceability21/25
    10 of 12 key fields link to a named source.
  • Direct evidence20/20
    4 direct source quotes retained.
  • Date provenance10/10
    The publication date comes from the source page or publisher.
  • Source transparency8/10
    1 source reference document publisher, URL, authority, retrieval, and supported fields.
  • Implementation specificity10/10
    Rewards source-backed components, data flow, quantified impact, and a specific outcome.
02

Traceability

Sources & provenance

Field coverage shows which facts each source supports. Dates are shown only when documented by the source.

  1. Baidu AI Cloud

    China Mars Exploration Project (中国火星探测工程) customer story

    First-party
    https://cloud.baidu.com/customer/case/zhurong.html
    Source type
    Vendor-published customer story
    Published
    May 15, 2025
    Supports
    • Customer
    • Business problem
    • Solution
    • Products
    • Solution components
    • Data flow
    • Integrations
    • technical area
    • use case category
    • Business outcome
    • outcome category
    • Metrics
    • Architecture
    • evidence quotes
    • Publication date

Date evidence

  • Published on source page May 15, 2025 · Baidu AI Cloud · First-party
    Open date source
Not separately sourced Industry, Region, Company size. Treat these normalized attributes as lower-confidence until a dedicated source is added.

Related

Similar use cases

See all like this
Baidu AI Cloud Oct 12, 2025 Guiyang Public Security Traffic Management Bureau (贵阳市公安交通管理局) Public Sector China Guiyang needed a more systematic way to reduce congestion in a mountainous city where road geometry, rapid growth, and fragmented traffic-management methods made it hard to optimize signal timing and respond to changing… Average vehicle delay at target intersections fell 15%. Real-time Analytics AI Traffic Signal Control Large Model · Qianfan Application Development Platform · Qianfan Model Development Platform Baidu AI Cloud Sep 24, 2025 All-China Federation of Trade Unions Public Sector China The All-China Federation of Trade Unions needed a unified AI foundation that could support legal aid, rights protection, and other worker services at national scale instead of leaving union service digitization… Delivery cycle fell from 3-6 months to under 1 month. Customer Service AutomationEnterprise Knowledge Search Qianfan AppBuilder · Qianfan ModelBuilder · Union Industry Model Baidu AI Cloud Jun 23, 2025 Shenzhen Institute of Artificial Intelligence and Robotics for Society (AIRS) Public Sector China AIRS needed scalable compute, storage and scheduling infrastructure to tackle the algorithm, data and compute bottlenecks that were limiting embodied-intelligence research and deployment. End-cloud coordination latency was reduced to within 50 ms. BaiGe AI Heterogeneous Compute Platform · GPU Cloud Servers · High-Performance Computing Platform