Cloud-edge-end collaboration for power line inspection: a proximal policy optimisation-based image offloading approach to minimize task execution time

Abstract The increasing complexity of power line infrastructure necessitates efficient inspection methods to ensure grid safety, with Unmanned Aerial Vehicles (UAVs) emerging as a key solution for automating image capture and fault detection. However, UAV-based systems face challenges such as limited onboard computational resources and high task execution latency, particularly when relying on centralized cloud processing. To address these issues, we propose a cloud-edge-end collaboration framework for power line inspection, where UAVs dynamically offload image processing tasks to edge or cloud servers based on real-time workload conditions. The offloading decisions are optimized using the Proximal Policy Optimization (PPO) algorithm, aiming to minimize task execution time while balancing computational demands across the architecture. Experimental results demonstrate that our approach reduces task execution time by up to 22% compared to traditional methods, offering a scalable and adaptive solution for efficient UAV-assisted power line inspections in smart grid applications. Beyond performance gains, this work contributes a generalizable framework for intelligent task scheduling in UAV systems, opening new possibilities for real-time, learning-based edge-cloud collaboration in smart infrastructure applications.