An Energy-Efficient Edge Ai Framework For Real-Time Object Detection in Mobile Robotics

Abstract

Mobile robots operating in dynamic environments require real-time object detection for navigation, obstacle avoidance, and autonomous decision-making. Traditional cloud-based Artificial Intelligence (AI) solutions suffer from communication latency, bandwidth limitations, and privacy concerns. Edge AI enables intelligent processing directly on embedded robotic platforms, reducing response time and improving reliability. However, limited computational resources and power constraints remain significant challenges. This research proposes an Energy-Efficient Edge AI Framework (EEEAIF) for real-time object detection in mobile robotics. The framework integrates lightweight deep learning models, adaptive model compression, dynamic voltage and frequency scaling (DVFS), and intelligent task scheduling to achieve high detection accuracy with reduced energy consumption. Experimental results demonstrate that the proposed framework achieves 95.2% detection accuracy while reducing energy consumption by 37% compared with conventional edge AI implementations. The proposed architecture provides a practical solution for autonomous robots operating in resource constrained environments.