MathWorks joins Edge AI group to boost embedded AI

MathWorks has joined the Edge AI Foundation, a non-profit group focused on energy-efficient artificial intelligence for edge devices, as it targets wider adoption of embedded AI in engineered systems.

The membership connects MathWorks with the foundation's network of supporters and contributors. The collaboration will focus on using MATLAB and Simulink to build AI models, integrate them into system simulations, and deploy them to embedded hardware.

Edge AI runs models on devices such as microcontrollers, FPGAs, and embedded GPUs, rather than relying on cloud infrastructure. The approach is promoted as a way to reduce latency, keep data local, and lower power consumption. The foundation, formerly the tinyML Foundation, positions itself as a forum for standardisation, education, and industry coordination around these goals.

Engineers often face constraints when moving AI workloads from servers to embedded devices, including limited compute, memory, and power. Teams also need ways to validate how AI components behave within wider systems that include sensors, controls, and safety requirements.

Workflow focus

MathWorks is best known for MATLAB, a programming environment widely used in engineering and science, and Simulink, a graphical environment for simulation and model-based design. It describes the tools as providing an end-to-end workflow for embedded AI, spanning training, integration, and deployment.

Simulink supports system-level simulation to test behaviour before software is deployed to target hardware. The tools also support verification and validation for safety- and mission-critical environments.

For deployment, MathWorks supports generating optimised C/C++, CUDA, and HDL code from the same Simulink model. It also cited compression techniques for resource-constrained devices and the ability to work across multiple AI frameworks, including MATLAB and PyTorch, as well as TensorFlow, ONNX, and XGBoost.

"MathWorks joining the EDGE AI FOUNDATION strengthens our shared mission to make edge AI more accessible," said Pete Bernard, Executive Director, EDGE AI FOUNDATION.

"As a recognised leader in embedded AI for engineered systems, MathWorks brings proven capabilities for AI model integration, system-level simulation, and optimised code generation. These contributions will be invaluable to our community as we work together to accelerate advancements in edge AI."

Industry use

MathWorks outlined examples of embedded AI work in automotive, aerospace, and industrial automation. In automotive engineering, MATLAB and Simulink can be used to create virtual sensors, such as estimates of battery state of charge or motor temperature. These models can then run in real time on microcontrollers in constrained environments.

In aerospace, teams develop anomaly detection and predictive maintenance algorithms that can be deployed on FPGAs. Latency and safety requirements in flight-critical systems are key drivers for this approach.

In industrial automation, the tools can be used to develop defect-detection algorithms for visual inspection, with deployments on embedded GPUs for high-speed quality control.

The Edge AI Foundation describes itself as a global community for efficient and scalable edge AI technologies. It says its network includes more than 100 Fortune 500 technology companies, and it cited engagement across online channels and participation in tinyML education programmes.

MathWorks has long had a strong footprint in engineering organisations that combine control systems, signal processing, and embedded software development. Interest in embedded AI has grown as these teams incorporate machine learning into products that must operate under strict constraints, including deterministic behaviour and safety assurance.

"Joining the EDGE AI FOUNDATION is a natural extension of our commitment to empowering engineers and scientists to innovate in AI, machine learning, and edge computing," said Lucas Garcia, Product Manager, AI, MathWorks.

"Our workflow enables teams to validate AI models developed in MATLAB and PyTorch through full-system simulation, optimise them for tight compute and memory constraints, and deploy across a wide spectrum of embedded hardware platforms," Garcia said.

"We look forward to collaborating with the Foundation and its members to advance the deployment of reliable, efficient AI solutions that address real-world challenges," he added.