TinyML in Industrial IoT: A Systematic Review of Applications, System Components, and Methodologies

Area of Science:

• Computer Science
• Electrical Engineering
• Industrial Engineering

Background:

• Tiny Machine Learning (TinyML) is vital for Industrial Internet of Things (IIoT) systems needing low latency and energy efficiency.
• Deploying TinyML in IIoT is complex due to diverse applications, hardware, and methodologies.
• Existing reviews lack a unified understanding of TinyML within IIoT contexts.

Purpose of the Study:

• To systematically review and synthesize TinyML applications, system components, and methodologies in IIoT.
• To address the gap in comprehensive literature on TinyML for IIoT.
• To provide a structured overview of current TinyML-enabled IIoT systems.

Main Methods:

• Systematic literature review (SLR) of 35 peer-reviewed studies (2018-2026).
• Synthesis of works across applications, system components, and methodologies.
• Comparative analysis of application categories based on accuracy, latency, memory, and energy.

Main Results:

• Primary TinyML applications in IIoT include predictive maintenance, equipment monitoring, anomaly detection, and energy management.
• Microcontroller-based hardware, lightweight software frameworks, and vibration sensing are dominant system components.
• Methodologies involve diverse data foundations, model selection, and optimization strategies.

Conclusions:

• This SLR consolidates diverse TinyML-IIoT aspects, including applications, hardware, software, sensing, and optimization.
• Identifies key challenges and outlines future research directions for TinyML in IIoT.
• Provides a structured foundation for understanding and advancing TinyML adoption in industrial settings.