Energy-Efficient Multi-Agent Deep Reinforcement Learning Task Offloading and Resource Allocation for UAV Edge...

Area of Science:

• Computer Science
• Artificial Intelligence
• Networking

Background:

• Mobile Edge Computing (MEC) systems face challenges with latency-sensitive applications.
• Unmanned Aerial Vehicles (UAVs) offer mobility and flexible deployment for MEC.
• Optimizing resource allocation and task offloading in UAV-MEC is crucial.

Purpose of the Study:

• To propose a novel multi-agent reinforcement learning framework for UAV-assisted MEC.
• To optimize task offloading and resource allocation for reduced latency and energy consumption.
• To address challenges in distributed decision-making and mobility-energy tradeoffs.

Main Methods:

• Developed the Multi-Agent Twin Delayed Deep Deterministic Policy Gradient for Task Offloading and Resource Allocation (MATD3-TORA) framework.
• Formulated a joint optimization problem to minimize latency and energy consumption.
• Enabled collaborative decision-making among multiple UAVs for efficient service provisioning.

Main Results:

• MATD3-TORA demonstrated significant improvements in system latency.
• The framework achieved enhanced energy efficiency compared to conventional methods.
• Validated effectiveness in real-time resource allocation and mobility-energy tradeoff management.

Conclusions:

• MATD3-TORA is an effective solution for optimizing UAV-assisted MEC networks.
• The proposed framework successfully addresses key challenges in distributed task offloading and resource management.
• Highlights the potential of multi-agent reinforcement learning in enhancing UAV-MEC performance.