On Device & Edge AI for Mobile

Abstract

On?device and edge artificial intelligence (AI) has emerged as a crucial technology in mobile computing, offering real?time intelligence, enhanced privacy, reduced latency, and decreased bandwidth consumption. With the proliferation of smartphones, wearables, and IoT devices, edge AI allows models to run locally or near the user rather than relying exclusively on centralized cloud systems. This paper surveys the current landscape of on?device and edge AI for mobile platforms, discusses key enabling technologies, design challenges, applications, performance trade?offs, and future research directions. We examine processor architectures, neural network optimization techniques, communication models, and privacy considerations. Through an assessment of existing frameworks and deployment strategies, we present insights on how edge AI reshapes mobile intelligence and makes AI more responsive and energy?efficient. The review concludes that while significant progress has been made, several challenges remain before truly ubiquitous, scalable, and secure edge?centric mobile AI systems become widespread.

KEYWORDS: On?device AI, Edge computing, Mobile intelligence, AI acceleration, Privacy, Neural network optimization, Low?power AI