AI-Based Test Case Generation from Requirements and MBSE Models

Overview

PoC #15 introduces the Requirement-to-Test Tool, an innovative solution designed to significantly reduce the manual effort and error susceptibility in test case definition. By analyzing relationships between requirements and integrating System Models (MBSE), the tool generates complex, executable test cases automatically.

Key Innovations

• Relational Requirement Analysis: The tool identifies "boundary conditions" within your requirement set. For example, it links a coolant pump's flow requirements with its specific energy constraints and fluid types to build a holistic test scenario.
• MBSE & RAG Integration: To enhance system understanding, the tool uses Retrieval Augmented Generation (RAG) to feed MBSE system models as context into the AI, ensuring the generated tests are architecturally sound.
• Few-Shot Learning: Users can toggle domain-specific examples to guide the AI’s output quality, ensuring high precision even in specialized automotive domains.

Workflow & Features

The tool provides a seamless end-to-end stream for engineers:

1. Clustering: Requirements are grouped by lifecycle phases and system states to optimize the processing frame for the Large Language Model (LLM).
2. Flexible Generation: Engineers can choose between Strict Generation (only using pre-defined parameters) or Extended Generation (allowing the AI to suggest logical boundary conditions).
3. BMW Keyword Export: A major highlight is the direct translation of natural language test steps into the official BMW Keyword Format, supporting Interior, Body, and specific Power-Train lists.

Impact

By automating the path from requirement to test script, KIMBA ensures that verification is not only faster but also more robust, directly contributing to the prevention of costly late-stage errors and vehicle recalls.