Proximal Policy Optimization-based Task Offloading Framework for Smart Disaster Monitoring using UAV-assisted WSNs.

Area of Science:

• Computer Science
• Artificial Intelligence
• Wireless Communication

Background:

• Unmanned Aerial Vehicles (UAVs) are vital for Wireless Sensor Networks (WSNs) in disaster monitoring.
• Key challenges include limited energy, dynamic task allocation, and trajectory optimization for UAVs.

Purpose of the Study:

• To propose a novel framework, ETORL-UAV, for energy-efficient task offloading in UAV-assisted WSNs.
• To enhance UAV operations in edge-enabled WSNs using reinforcement learning.

Main Methods:

• Developed ETORL-UAV, integrating Proximal Policy Optimization (PPO) based reinforcement learning.
• Implemented a multi-objective reward model to balance energy, task success, and network lifetime.

Main Results:

• ETORL-UAV achieved up to 9.3% higher task offloading success.
• Demonstrated an 18.75% improvement in network lifetime and a 27% reduction in energy consumption.
• Outperformed five state-of-the-art methods in simulations.

Conclusions:

• ETORL-UAV offers a scalable and reliable solution for UAV-assisted WSNs.
• The framework is practical for real-world disaster-response deployments.
• Reinforcement learning effectively optimizes UAV operations for WSNs.