Developing Engineered Systems with AI and MATLAB

This work presents the development of an AI-based Reduced Order Model (ROM) derived from a high-fidelity battery thermal-electrical model in AVL CRUISE™ M. Created using the MathWorks Reduced Order Modeling Support Package, the ROM predicts state-of-charge (SOC), terminal voltage, and temperature evolution under diverse operating conditions. Achieving over 100× real-time performance, it enables rapid what-if analyses and seamless integration into HiL platforms. The methodology combines domain-specific feature engineering with nonlinear autoregressive neural networks to accurately capture transient behavior while maintaining robustness across varying current profiles, coolant flow rates, and thermal boundary conditions.

This work represents the first stage in a multi-year initiative to apply AI-ROM techniques to multiple vehicle subsystems. The long-term vision is to integrate these subsystem ROMs into a connected System-of-Systems framework, enabling fast, high-fidelity simulations for predictive energy and thermal management strategies. This scalable approach bridges physics-based modeling with AI acceleration, supporting optimization-driven integration and cross-domain trade-off studies toward sustainable mobility.

Working with automotive MF4 log files at scale is a growing challenge, especially since there’s no straightforward way to directly consume .mf4 data in Spark or Databricks. For teams who rely on trusted MATLAB workflows, this has meant painful workarounds or missed opportunities to analyze valuable field data in the cloud. In this session, see how MATLAB’s Vehicle Network Toolbox lets you design and test MF4 data workflows locally, then easily package and deploy your MATLAB code to Databricks for scalable, cloud-based analysis. Seamlessly take your trusted MATLAB and Simulink workflows from desktop to cloud, unlocking powerful new possibilities for automotive analytics.

Session Highlights:

• Connect MATLAB with Databricks 
• Develop and testing MATLAB functions locally 
• Deploy MATLAB code as a Python .whl file to Databricks 
• Process Databricks data using MATLAB workflows

Are you ready to break through the barriers of automotive data analysis? Join us for this exciting session to see how the MATLAB Interface for Databricks enables you to deploy, run, and scale your MATLAB analyses in the cloud, using the tools and code you already know and trust.

As companies become increasingly data-driven, Simulink users are being challenged to use real-world data to inform and validate their designs. You already have powerful data analysis and analytics capabilities at your fingertips with MATLAB. Discover how to unlock the full potential of MATLAB within your Simulink workflows and learn practical techniques to meet the data-driven demands of modern engineering, all without leaving your existing toolchain. These approaches can boost the robustness and reliability of your simulations, and open doors to advanced applications like Digital Twins.

Session Highlights:

• Ingesting data from log files
• Programmatically configuring Simulink model parameters
• Simulating in serial or parallel
• Efficiently post-processing simulation results

Whether you want to accelerate your workflow, minimize errors, or break into new frontiers, this session will take your Simulink projects to the next level.

Bridging skills gaps and staying ahead in the rapidly evolving automotive sector requires continuous upskilling and access to the right resources. MathWorks Training Services empowers teams to accelerate innovation by offering a comprehensive suite of learning solutions, ranging from flexible online courses to in-depth, instructor-led sessions, tailored to the unique challenges of digital engineering and AI in automotive applications.

In this session, discover how our training services can help your team master essential skills, adopt new technologies, and drive measurable performance improvements. Explore new opportunities for upskilling and learn how we can partner with you to align training with your strategic goals.

Session Highlights:

• Tailored learning paths for automotive engineering and AI
• Strategies for closing skills gaps and boosting team productivity
• Opportunities for partnership and customized training solutions

Join us to see how MathWorks Training Services can help you build the expertise needed to deliver on your most ambitious automotive projects.

MATLAB and Simulink with deep learning frameworks, TensorFlow and PyTorch, provide enhanced capabilities for building and training your machine learning models. Via interoperability, you can take full advantage of the MATLAB ecosystem and integrate it with resources developed by the open-source community. You can combine workflows that include data-centric preprocessing, model tuning, model compression, model integration, and automatic code generation with models developed outside of MATLAB.

Analyzing data through scripts or standalone applications is common, but moving data analytics to the web provides greater flexibility for accessing and processing both logged and streaming data. In this session, you will learn about developing interactive data analytics applications in MATLAB and deploy them as web apps using MATLAB Web App Server. These web apps feature user-friendly, browser-based interfaces, enabling users, even those without MATLAB experience to interact with and perform analytics. Finally, discover how to further extend these analytics capabilities by deploying them to the cloud with MATLAB Production Server, making them accessible and integrable with any production IT system.

Session highlights:

• Develop Interactive Applications: Learn to build interactive data analytics applications using MATLAB’s App Designer
• Web-based Deployment: Discover how to deploy MATLAB applications as web apps via MATLAB Web App Server
• Cloud-based Analytics: Explore how to outsource data analytics to the cloud using MATLAB Production Server
• Data-Compute Co-location: Understand the benefits of co-locating your compute resources with your data for enhanced performance and scalability

GenAI is becoming a key technology for engineers doing research and development. While the capabilities of the technology continue to expand, practitioners are learning how to best integrate GenAI with their workflows and get the most value from the tools. MathWorks is investing in copilots that enhance tools like MATLAB and Simulink, combining the benefits of high-level tools for algorithm development and model-based design with the speed and productivity gains of AI assistants.

In this talk we will discuss how these tools will impact R&D engineering workflows across a variety of use-cases including explaining code and models, advising engineers on tools and techniques, creating and improving code and models, verifying results, and automating workflows.

High-fidelity models—such as those based on FEA, CFD, and CAE—are indispensable for detailed component design but often too computationally expensive for system-level simulation, control design, or hardware-in-the-loop testing. Tasks like evaluating vehicle control systems or predicting emissions in real time cannot rely on models that take hours or days per run.

Reduced-order modeling (ROM) addresses this challenge by creating fast, lower-fidelity approximations that preserve essential system dynamics. Instead of restarting from scratch, engineers can reuse existing high-fidelity data to build AI-based surrogate models that accelerate analysis and enable real-time applications.