A Novel Software Framework for Optimizing Edge AI Systems: Overcoming Computational and Deployment Challenges

The rapid adoption of Edge Artificial Intelligence (Edge AI) in autonomous systems, smart cities, and IoT applications has introduced significant software engineering challenges related to computational efficiency, real-time processing, and scalable deployment. Traditional AI deployment models struggle with the resource constraints, energy efficiency, and heterogeneity of edge devices, limiting their ability to perform high-performance inference in real-world environments. In this research, we propose a novel software framework designed to optimize Edge AI model execution, adaptive deployment, and resource allocation across diverse hardware architectures. Our approach leverages lightweight AI model transformations, dynamic resource scheduling, and federated software updates to enhance the efficiency and scalability of AI-powered edge systems. We develop and test our framework on heterogeneous edge environments, demonstrating a ۳۰-۵۰% reduction in inference latency and power consumption compared to existing deployment techniques. Additionally, we introduce an adaptive learning mechanism that enables efficient model updates without requiring cloud dependencies. This research provides a scalable and secure software foundation for next-generation Edge AI applications, addressing critical limitations in computational overhead, real-time adaptability, and security.

کلیدواژه ها:

Edge AI ، Software Optimization ، Model Deployment ، Resource Allocation ، Federated Learning