Machine Learning for Electric Motor Drives: A Survey with Perspective on Deployment at the Edge

This review paper presents a systematic and comprehensive survey of the application of machine learning (ML) techniques in electric motor drives, with a particular focus on their practical deployment on embedded platforms and edge devices in motor control applications. The survey not only examines how ML methods are employed to realize partial to complete control strategies in motor drives but also provides an extensive overview of the growing ecosystem of hardware platforms suitable for ML deployment at the edge. Emphasis is placed on the growing availability and diversity of platforms that enable real-time, low-latency, and energy-efficient ML inference. Furthermore, the paper highlights the emerging potential of reinforcement learning (RL) as a promising approach for adaptive and intelligent motor drive control. Key interdisciplinary topics are thoroughly examined, including the fundamentals of motor control theory, diverse machine learning methodologies such as supervised and reinforcement learning for model training, advanced techniques for model compression and quantization to enable deployment on resource-constrained edge devices, as well as data preprocessing and feature extraction strategies critical for effective analysis and real-time inference in motor drive applications. Recent advancements in embedded tools and frameworks are also highlighted, assisting researchers and engineers in leveraging ML at the edge for practical motor control solutions.