AI-Powered Requirements: Automating MBSE for Smarter Automotive Specs

AI powered requirements with MBSE

Welcome back to MBSE Explained! In my years working as a model-based developer and architect in the automotive world, I’ve seen the complexity of vehicle systems explode. We’ve moved from isolated ECUs to a world of software-defined vehicles (SDVs) that are constantly connected, updated, and evolving. This shift has turned traditional requirements engineering—often managed in spreadsheets and documents—into a significant bottleneck. It's slow, prone to error, and struggles to keep up with the pace of innovation demanded by modern EVs and connected cars.

But what if we could automate the most tedious parts of this process? What if we could leverage artificial intelligence to help us create clearer, more consistent, and more robust requirements directly within our models? That’s exactly what the fusion of AI and Model-Based Systems Engineering (MBSE) promises. This isn't science fiction; it's the next logical step in building smarter systems. In this post, we'll explore how AI is set to revolutionize requirements engineering for the automotive industry, making our development cycles faster and our vehicles smarter.

The Modern Automotive Challenge: A Flood of Requirements

The software-defined vehicle isn't just a car with a big screen; it's a complex, distributed computing system on wheels. Consider the sheer volume and variety of requirements we now juggle:

• Connectivity: Features like Vehicle-to-Everything (V2X) communication and seamless cloud integration for telematics and over-the-air (OTA) updates.
• Functional Safety (ISO 26262): Strict mandates to ensure electronic systems don't fail in dangerous ways.
• Cybersecurity (ISO/SAE 21434): Protecting the vehicle from malicious attacks, a non-negotiable for connected systems.
• Customer Features: A constantly growing list of demands for infotainment, driver assistance (ADAS), and personalization.

Managing these interconnected requirements in isolated documents is a recipe for disaster. It leads to ambiguity, missed dependencies, and costly rework late in the development cycle. MBSE provides the structured, model-based foundation to manage this complexity, but the initial effort of capturing and refining thousands of requirements is still immense.

Where MBSE Meets AI: The Automation Powerhouse

This is where AI changes the game. Think of MBSE as the perfect chassis for systems engineering—it provides the structure, traceability, and a single source of truth. AI is the intelligent engine we can now place inside that chassis.

By integrating AI, specifically Natural Language Processing (NLP) and machine learning algorithms, into an MBSE workflow, we can automate tasks that have always been manual, time-consuming, and error-prone. The AI isn’t replacing the systems engineer; it’s becoming their most powerful assistant, capable of processing vast amounts of information and identifying patterns humans might miss.

Practical AI Applications for Automotive Requirements

So, how does this work in practice? Let's break down a few key applications that are already becoming a reality.

1. Extracting Requirements from Unstructured Sources

Engineers receive input from countless sources: customer feature requests, marketing documents, federal regulations, and technical standards. An AI-powered tool can scan these documents and perform a first pass at identifying and formalizing potential requirements.

• Example: An NLP model could parse a 300-page UNECE regulation on cybersecurity and automatically generate a set of candidate requirements, classifying them by priority and linking them to the source text. This alone can save weeks of manual work.

2. Automated Requirement Generation and Refinement

Once high-level stakeholder needs are in the model, AI can help decompose them into detailed, verifiable system and software requirements. It can learn from past projects and established patterns to ensure requirements are written clearly and consistently.

• Example: A high-level need like, "The vehicle shall support OTA updates for the powertrain controller," can be automatically broken down by an AI into specific requirements for security (e.g., "The update package shall be cryptographically signed"), performance (e.g., "The update shall complete within 15 minutes with the vehicle in a safe state"), and reliability (e.g., "A robust rollback mechanism shall be implemented in case of update failure").

3. Consistency and Conflict Detection

In a model with thousands of requirements, it's nearly impossible for a human to spot every contradiction. AI excels at this. It can analyze the entire requirements set within the MBSE model to flag potential issues.

• Example: The model might contain one requirement for a sub-second boot time for an infotainment ECU and another from a cybersecurity stakeholder demanding extensive boot-time integrity checks. An AI can immediately flag this conflict, allowing architects to negotiate a trade-off early in the design phase, not during integration testing.

4. Automating Traceability

Manually creating links between requirements, architecture blocks, and test cases is one of the most tedious tasks in MBSE. AI can suggest and even automatically create these traceability links based on semantic analysis of the text. This not only saves time but also ensures the model remains a true, interconnected digital thread.

Real-World Example: A Cloud-Linked EV Charging System

Let's ground this with a concrete automotive example. Imagine we are designing an onboard charger ECU that communicates with a cloud backend to authorize payments and optimize charging schedules.

1. Input: The high-level feature is "The ECU must securely manage charging sessions via the cloud." We also feed the AI relevant standards like ISO 15118 (for vehicle-to-grid communication) and ISO/SAE 21434 (cybersecurity).
2. AI + MBSE Process:
   • The AI's NLP engine parses the standards and identifies key mandates for authentication, encryption, and data privacy.
   • It auto-generates a series of formal requirements in SysML within the model, such as: "The ECU shall establish a TLS 1.3 encrypted channel with the cloud server before transmitting any user data."
   • It analyzes these new requirements against existing architectural constraints for the ECU, like its limited processing power or real-time operating system (RTOS) task deadlines.
   • It flags a potential conflict: the computational overhead of TLS 1.3 might impact the ECU's ability to meet a hard real-time deadline related to battery monitoring.
3. Outcome: The systems architect is immediately alerted to a critical design trade-off. Instead of discovering this during hardware-in-the-loop testing, the team can address it upfront—perhaps by selecting a more powerful microprocessor or optimizing the security protocol implementation.

The Road Ahead

AI is not a magic wand. The quality of its output depends heavily on the quality of the training data and the clarity of the input documents. Human oversight from experienced domain experts is—and will remain—absolutely critical. The engineer's role will shift from manually writing every requirement to validating, refining, and curating the output generated by their AI assistant.

This partnership between human intellect and artificial intelligence is the future of complex systems design. By automating the mundane, we free up engineers to focus on what they do best: innovation and problem-solving.

Conclusion

Integrating AI into our MBSE workflows is no longer a futuristic concept; it's a practical solution to the overwhelming complexity of modern automotive systems. It allows us to build better, safer, and more reliable software-defined vehicles by ensuring our requirements are complete, consistent, and traceable from day one. By embracing this technology, we can accelerate development, reduce errors, and deliver the cutting-edge features that customers demand.

What are your thoughts on using AI for requirements engineering? Have you experimented with any tools in this space? Share your experience in the comments below! Let's continue the conversation on simplifying systems for smarter EVs.