Two-Layer Research Framework
We investigate Intelligent Model-Driven Engineering as both a software engineering foundation and an application framework for intelligent software factories.
Research Focus
Our central question is how to combine model-driven precision and data-driven intelligence so that software development becomes high-quality, high-efficiency, transparent, traceable, and suitable for complex systems.
Layer 01
We study Intelligent Model-Driven as a software engineering foundation: models provide the precise and verifiable structure, while generative AI and agent workflows provide the adaptive intelligence for building, refining, and connecting engineering artifacts.
Layer 02
We turn the engineering theory into intelligent software factories that support requirements understanding, model construction, architecture and design generation, code generation, deployment, and feedback-driven evolution across the software lifecycle.
Layer 01
This layer studies the concepts, languages, transformations, constraints, agent workflows, and generation mechanisms that make intelligent software production controllable and verifiable.
We define requirements models, design models, deployment models, graphical DSLs, textual DSLs, metamodels, and contracts as the precise carriers of software knowledge.
We study how large language models and agents can generate, repair, optimize, and explain models and code while staying constrained by explicit engineering structures.
We combine model-driven precision with data-driven flexibility to make software generation efficient, controllable, traceable, and suitable for complex and critical systems.
We investigate how to decompose complex software development into model-centered steps and connect them into a lifecycle chain from requirements to deployable systems.
Layer 02
We build software production lines where requirements, models, agents, generators, tests, release scripts, and deployment environments are assembled into repeatable and intelligent development workflows.
A software factory assembles tools, models, workflows, scripts, environments, and quality gates into automated production lines for software development, build, test, release, and delivery.
RM2PT is studied as a full-lifecycle intelligent software factory platform for automatic requirements extraction and refinement, validation, architecture generation, design optimization, code generation, and deployment.
We extend the approach to domain software factories such as HarmonyOS application generation, where GUI models, domain models, and AI-assisted code generation improve efficiency and quality.
01
Requirements model
02
Prototype validation
03
Architecture model
04
Design model
05
Code generation
06
Deployment and feedback
Research Progress
The research has been evaluated through full-lifecycle RM2PT workflows and domain scenarios such as HarmonyOS application generation, where model-driven structures and AI generation are integrated.
~13x
faster requirements extraction and confirmation
2-10x
higher design productivity in evaluated workflows
61.92%
coding efficiency improvement in RM2PT studies
98%+
code correctness in HarmonyOS case validation
Capability Portfolio
The following projects are organized as reusable capability modules for Intelligent Model-Driven Engineering applications and intelligent software factories.
From natural language, user stories, use cases, goals, problem frames, SysML, and UML to analyzable requirements and design models.
From OCL, formal contracts, validation scripts, generated inputs, and traceability evidence to quality-aware generation.
From requirements models to monolithic systems, microservices, enterprise information systems, DApps, GUI code, and domain applications.
Automated Requirement Elicitation
Model-driven tool support for intelligent software engineering workflows.
Model-driven tool support for intelligent software engineering workflows.
Model-driven tool support for intelligent software engineering workflows.
Model-driven tool support for intelligent software engineering workflows.
Model-driven tool support for intelligent software engineering workflows.
Automatic Generation SRS from Requirements Models
Automated Generation of UML from iStar Model
Problem Frames to UML
Automated Generation of Prototype Input for Rapid Requirements Validation
ChatGPT-enhanced Initialization Data Generation for Rapid Requirements Validation
A Tool for Automatic Generation of Validation Scripts to Support Rapid Requirements Validation
Automatic Generation of Natural Language from Object Constraint Language
Interactive Generation of Object Constraint Language from Natural Language
Object Constraint Language Generation from NL
Automatic Generation of Monolith Design from Requirements Model
Microservice Architecture Generation and Refinement from Requirements Model
Automatic Generation of MicroServices Architecture from Requirements Model
Automatic Generation of Enterprise Information System (SpringBoot-based Backend)
Automatic Generation of Enterprise Information System (Golang-based BackEnd)
Automatic Generation of Blockchain-based System (DApp)
Automatic Generation of Frontend GUI from Requirements Model