Our Research

Trustworthy Intelligent Model-Driven Engineering

Trustworthy Intelligent Model-Driven Engineering integrates models and AI in both directions: AI supports the construction, generation, and evaluation of models and domain-specific languages; models guide the definition and execution of agents and engineering workflows; and explicit semantics, contracts, refinement checks, verification, and traceability provide assurance for property-preserving model evolution and its outcomes.

Models form the engineering backbone.

Models · The main engineering axis

AI · Connects and accelerates work along the models

Formal Methods

Property-preserving refinement across stages

Models form the engineering backbone.Models are the main axis of the engineering process: they express intent, structure, behavior, and evidence across requirements, design, realization, and verification and validation. AI works along this model backbone, helping construct, connect, transform, and evaluate the models. Formal methods check both model properties and the refinement relations between models, assessing whether established intent, constraints, and required properties are preserved as detail increases. Realization may involve implementation, integration, manufacturing, or preparation. Verification checks conformance to specified requirements and properties; validation assesses fitness for intended use. Both occur throughout the process with feedback and iteration. Paths may branch, and constituent systems may evolve independently.RequirementsEngineering modelIntent & propertiesUnderstand & assessDesignEngineering modelStructure & behaviorConnect & transformRealizationEngineering modelDetailed realizationGenerate & evolveVerification &ValidationEngineering modelOutcomes & evidenceVerify & validateRefinement ↓Semantic consistencyRefinement ↓Preserve constraintsRefinement ↓Preserve properties
Consistency checksConstraint preservationProperty verification
Models are the main axis of the engineering process: they express intent, structure, behavior, and evidence across requirements, design, realization, and verification and validation. AI works along this model backbone, helping construct, connect, transform, and evaluate the models. Formal methods check both model properties and the refinement relations between models, assessing whether established intent, constraints, and required properties are preserved as detail increases. Realization may involve implementation, integration, manufacturing, or preparation. Verification checks conformance to specified requirements and properties; validation assesses fitness for intended use. Both occur throughout the process with feedback and iteration. Paths may branch, and constituent systems may evolve independently.

From foundations to industrial impact

Models structure engineering knowledge, AI supports generation and refinement, and formal methods provide constraints and verification. Industrial problems guide the development and evaluation of our theories, methods, and tools.

01 · 核心基础研究

Core Foundations

MDA · Intelligent MDA · Trustworthy Intelligent MDA

02 · 工程方法研究

Engineering Methodologies

Intelligent Software Engineering · Intelligent Systems Engineering · Intelligent System-of-Systems Engineering

03 · 工业领域应用

Industrial Applications

Intelligent HarmonyOS Development · Intelligent Manufacturing · Intelligent Pharmaceutics and Health

01 · 核心基础研究

Core Foundations

Metamodels, domain-specific languages, model evaluation, and transformations form the engineering foundations. AI enhances these methods, models define and guide AI agents, and formal methods support trustworthy collaboration between the two.

模型驱动架构

MDA

Modeling Is Programming建模即编程

Research objective: Model Definition解决的问题：模型可定义

We study the foundations of model-driven architecture (MDA): domain-specific language (DSL) design and evaluation, automated model transformation, and model-based code generation. Research connects language definitions and modeling semantics with transformation rules and generated implementations, and evaluates language expressiveness, transformation correctness, and the quality of generated code.

Research team

Alumni contributors

智能模型驱动架构

Intelligent MDA

Intelligent Modeling智能建模

Research objective: Model Generation解决的问题：模型可生成

We study the bidirectional relationship between AI and model-driven architecture. AI for MDA focuses on intelligent model design, generation, and evaluation. MDA for AI focuses on the precise definition and generation of AI system elements, and the development of AI systems using those elements.

Research team

可信智能模型驱动架构

Trustworthy Intelligent MDA

Trustworthy Intelligent Modeling可信智能建模

Research objective: Model Verification解决的问题：模型可验证

We study the verifiability and consistent propagation of model semantics across intelligent model design, generation, and transformation.

Research team

Alumni contributors

02 · 工程方法研究

Engineering Methodologies

Lifecycle methods and toolchains for software, systems, and systems of systems.

智能软件工程

Intelligent Software Engineering

We study requirements confirmation through executable prototypes and selected behavior demonstrations, architecture and design refinement, frontend and backend generation, design optimization, and model extraction from existing code.

Related work includes discovering reusable agent services from partially specified tasks and using domain knowledge to support software generation, including government-domain applications.

Research team

智能系统工程

Intelligent Systems Engineering

We study requirements decomposition and traceability, SysML-based structure and behavior modeling, and AI-assisted modeling tools for complex systems.

Probabilistic reliability research examines how AI behavior uncertainty propagates through a system, how system-level reliability requirements can be decomposed, and when the resulting failure probability can be bounded and assessed. Engineering analysis and simulation provide evaluation evidence.

Research team

智能体系工程

Intelligent System-of-Systems Engineering

We study mission-oriented system-of-systems engineering: decomposing high-level missions, expressing constituent systems, capabilities, resources, and coordination constraints, and bridging mission models to simulation. Simulation evaluates whether the coordinated sub-missions support the overall mission goal.

Research team

03 · 工业领域应用

Industrial Applications

Representative applications include HarmonyOS development, intelligent manufacturing, and pharmaceutics and health. Engineering methods and AI capabilities can support more than one domain; additional cases include aviation requirements and government-domain software generation.

智能鸿蒙开发

Intelligent HarmonyOS Development

Intelligent software engineering supports model-based HarmonyOS application development. Current work focuses on GUI refinement and ArkTS generation, with integration into an application-development workflow. Representative outcomes include AutoArk.

Research team

智能制造

Intelligent Manufacturing

We study model-driven transfer across heterogeneous CAD, CAE, and CAM tools, supported by distributed task execution for multidisciplinary simulation.

A separate research topic aligns and fuses existing multimodal production-process data for a specific analysis task. Engineering models and production data retain their distinct roles.

Research team

智能药剂与健康

Intelligent Pharmaceutics and Health

We study domain-knowledge-enhanced AI-agent methods for in-vitro pharmaceutical formulation performance prediction, building on existing deep-learning work. Evaluation covers selected prediction tasks, compares against established methods, and examines how domain knowledge contributes to performance.

Our health-related research also includes AI methods for electronic health records (EHR).

Research team

Research Projects

  • Microservice Design for Complex Software Systems
  • Model-Driven Software Modeling for Automotive Electronic Control Units
  • UAF-Based Mission Architecture Modeling Language
Explore all research projects

Research Papers

View all selected publications