Journal of Embedded Systems & Its Applications ISSN: 3139-2296 (Online)

Authors:- Prabhat Sharma, Devansh Kulkarni, Meenal Joseph, Arvind Jaiswal

Authors:-Edge Artificial Intelligence (Edge AI) is becoming an important paradigm where AI models are deployed directly on embedded devices instead of relying fully on cloud computing. This approach is very useful for real-time decision making, reduced latency, better privacy and lower bandwidth usage. However, embedded platforms such as microcontrollers, single board computers, and IoT devices usually have very limited computational power, memory, and energy resources. Running complex AI models on such platforms becomes a major technical challenge. This paper presents a review of techniques, architectures, optimization methods, and tools that enable deployment of Edge AI on resource-constrained embedded systems. It discusses model compression, quantization, pruning, hardware accelerators, and lightweight neural network architectures. The paper also highlights practical use cases, current challenges, and future research directions in this emerging field.

Keywords Edge AI, Embedded Systems, TinyML, Model Compression, Low-Power Devices, Quantization, IoT, Neural Networks