MathWorks joins the Edge AI Foundation, strengthening the development flow for designing and verifying artificial intelligence in embedded systems. The aim is to go beyond simply building AI models and connect this process to system-level performance testing before deployment on actual equipment, as well as deployment tailored to limited hardware environments.

MathWorks announced on April 13 that it has joined the Edge AI Foundation, a non-profit organization promoting the spread of energy-efficient AI technology for edge devices. Accordingly, MathWorks plans to support embedded AI development environments utilizing MATLAB and Simulink by collaborating with the Foundation's global network.

The core of this collaboration lies in consolidating the entire process of integrating AI into engineering systems into a single workflow. MathWorks presented a structure that spans AI model training and integration, system-level simulation, optimized code generation, and embedded device deployment. It also includes pre-deployment behavior verification, C, C++, CUDA, and HDL code generation, and model compression features for resource-constrained devices.

In addition, MATLAB and Simulink integrate with various AI frameworks such as PyTorch, TensorFlow, ONNX, and XGBoost, and provide low-code apps that allow even users with limited professional AI development experience to attempt model training. It is also explained that verification and validation functions are supported for systems where safety and reliability are critical.

Application examples were also presented. In the automotive sector, virtual sensors that estimate battery charge status or motor temperature are developed and deployed on microcontrollers, while in the aerospace sector, the development of FPGA-based anomaly detection and predictive maintenance algorithms is supported to suit environments with strict latency and safety requirements. In industrial automation, a method for developing visual inspection-based defect detection algorithms and deploying them on embedded GPUs was introduced.

Since Edge AI must process data directly on field equipment rather than in the cloud, constraints such as computational resources, power consumption, and latency must be taken into account. MathWorks' entry is interpreted as an example of the industry's trend to address not only AI model performance but also the verification of actual system operations and hardware deployment feasibility. In the engineering field, competition for related tools is expected to continue in a direction that bridges the gap between AI development and system verification.