Page 805 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

Use Case 5: Intelligent In-Vehicle Scenario Orchestration Agent

System Transport 4.10: Intelligent

Country: China

Organization: Chongqing Changan Automobile Co. Ltd.

Contact Person: Yonggang Luo(luoyg3@ changan .com .cn)

1 Use Case Summary Table

Item Details
Category Intelligent transport

Problem Addressed Traditional vehicle control configuration processes are cumbersome, frag-
mented, and heavily reliant on professional knowledge, requiring users
to invest significant time and effort to complete tasks.

Key Aspects of Solu- The system leverages large language models to accurately recognize
tion and understand user intentions. It automatically generates logical chains
containing “preconditions”, “triggers”, and “executions”, significantly
improving scenario orchestration efficiency and scalability.

Technology Large Language Models, Service Orchestration.
Keywords
Data Availability Private

Metadata (Type of Text
Data) Input data: Chinese user intents
Output data: Chinese and English car services

Model Training and SFT With approximately 20 million tokens of automotive domain data and
Fine-Tuning around 400,000 instruction tuning data.

Testbeds or Pilot TikTok Video [1]
Deployments

Code repositories None

2 Use Case Description

2�1 Description

Intended use: The system allows users to dynamically create and manage complex in-vehicle
scenario modes, such as personalized climate control and entertainment settings, using natural

769