Artificial intelligence in embedded systems: A review of techniques, applications, and challenges

Abstract:

The convergence of Artificial Intelligence (AI) and embedded systems has revolutionized modern electronic engineering, enabling intelligent functionalities in devices with constrained power, memory, and processing resources. This review explores the evolution, techniques, hardware platforms, and application domains of embedded AI, focusing on advancements such as TinyML, federated learning, and hybrid models. It further categorizes the landscape of AI-capable embedded hardware, including microcontrollers, SoCs, FPGAs, ASICs, and modular AI accelerators. Real-world deployments across agriculture, healthcare, automotive, IIoT, robotics, and smart cities are discussed, with emphasis on privacy-aware, real-time, and energy-efficient implementations. The paper also outlines critical challenges such as computational limits, latency, model updates, and security risks. Lastly, it highlights emerging trends including neuromorphic computing, self-learning models, cross-platform ML deployment, and hardware-algorithm co-design, offering a forward-looking perspective on the future of AI in embedded applications.