Task Offloading Strategy for Unmanned Aerial Vehicle Power Inspection Based on Deep Reinforcement Learning

Area of Science:

• Electrical Engineering
• Computer Science
• Artificial Intelligence

Background:

• Traditional power inspection methods are inefficient and risky.
• Unmanned aerial vehicles (UAVs) offer improved efficiency but have computational and energy limitations.
• Existing solutions struggle with computationally intensive and latency-sensitive inspection tasks.

Purpose of the Study:

• To develop an efficient UAV task offloading strategy for power inspection using deep reinforcement learning (DRL).
• To address the computational and energy constraints of UAVs in power grid monitoring.
• To optimize task processing latency and energy consumption in UAV-based power facility inspections.

Main Methods:

• Proposed a collaborative computing architecture integrating UAVs with mobile edge computing (MEC) servers.
• Developed a computational model for energy consumption and task processing latency.
• Formalized the task offloading problem as a multi-objective optimization and Markov Decision Process (MDP).
• Introduced an Optimized Task Offloading algorithm based on Deep Deterministic Policy Gradient (OTDDPG).

Main Results:

• The proposed UAV-Edge server collaborative computing architecture effectively leverages UAV mobility and MEC capabilities.
• The OTDDPG algorithm achieved significant reductions in task processing latency.
• The strategy demonstrated substantial improvements in energy consumption compared to baseline methods.
• Simulation results validated the effectiveness of the DRL-based task offloading approach.

Conclusions:

• The DRL-based task offloading strategy significantly enhances the efficiency and reduces the energy consumption of UAVs in power inspection.
• The developed collaborative computing architecture and OTDDPG algorithm provide an effective solution for UAV limitations in power grid monitoring.
• This approach offers a promising direction for intelligent and automated power facility inspection systems.