计算机科学
弹道
智能电网
实时计算
工程类
嵌入式系统
移动机器人
控制工程
智能传感器
钥匙(锁)
自动化
控制系统
网格
全球定位系统
加速度
遥控水下航行器
作者
Liang Zhao,Jun Li,Xiaojun Liu,Bibo Xiao,Junpeng Shi,Fangqing Wen,Jing Luo,Shouzhi Xu
标识
DOI:10.1109/tce.2026.3727995
摘要
Unmanned aerial vehicles (UAVs) play a significant role in the inspection and maintenance of smart grids. In general, an inspection UAV needs to sequentially visit multiple inspection points, capture inspection images, and perform onboard edge computing of these images to enable rapid fault diagnosis. However, due to its limited computation power and battery capacity, relying solely on UAV-based edge computing may result in higher latency, greater energy consumption, or even service interruptions. To address this challenge, we propose an edge-cloud collaborative UAV inspection system, where the UAV partially offloads its tasks to a cloud monitoring center deployed on a High-Altitude Platform (HAP), thereby mitigating onboard computational bottlenecks. In addition, several ground wireless charging stations provide energy replenishment for the UAV to sustain stable operation. Under task-deadline and energy-budget constraints, we aim to minimize the total inspection time by jointly optimizing the discrete visiting order and the continuous offloading and resource allocation decisions. For this non-convex mixed-integer nonlinear programming (MINLP) problem, we propose a two-stage TGTD optimization framework, which integrates heuristic search and deep reinforcement learning. In the first stage of TGTD, we propose a task-aware GA-based path planning method (TGP2) to obtain a near-optimal visiting order of inspection points. In the second stage of TGTD, we adopt Twin Delayed Deep Deterministic Policy Gradient (TD3) to solve the task offloading and UAV computing resource allocation problem. Simulation results demonstrate that the proposed TGTD method significantly reduces the inspection travel distance and total inspection time across diverse system configurations, which provides theoretical insights for practical UAV inspection applications.
科研通智能强力驱动
Strongly Powered by AbleSci AI