Exploring Agentic AI in Model-Based Design with MATLAB and Simulink

When building a virtual vehicle model, agentic AI enables a new way to accelerate system-level modeling by allowing AI agents to work directly inside MATLAB^® and Simulink^®. This presentation shows how agentic AI can speed up the build, test, and simulation of component models by using the Model Context Protocol (MCP) to let agents interact directly with engineering artifacts.

Building virtual vehicle models is highly iterative. Engineers review specifications and test data, determine how to model dynamics, and develop Simulink component models that must integrate into larger system simulations. Even with mature tools, this work involves frequent context switching across documents, scripts, models, and test results, which can slow development cycles. This presentation shows how an AI agent navigates Simulink hierarchies, updates plant and controller parameters, runs model-in-the-loop (MIL) tests, diagnoses errors, and summarizes results. The automotive examples demonstrate how the agent supports plant initialization, parameter sweeps, test harness execution, and documentation within a virtual test architecture.

Common workflow bottlenecks such as datatype mismatches and MIL test failures are handled autonomously, with engineers staying in the loop at key decision points to validate model behavior and system integration. Results show reduced integration time and rework, while maintaining traceability and control through human-in-the-loop checkpoints.