Multi-UAV-Assisted Task Offloading and Trajectory Optimization for Edge Computing via NOMA.

Area of Science:

• Wireless Communication
• Mobile Edge Computing (MEC)
• Network Optimization

Background:

• Unmanned Aerial Vehicles (UAVs) offer flexible deployment for enhancing Mobile Edge Computing (MEC) systems.
• Task queuing and network load imbalance in MEC increase user waiting delays.
• Existing solutions struggle with dynamic task loads and uneven service distribution.

Purpose of the Study:

• To propose a collaborative multi-UAV MEC network model to mitigate transmission queuing and load imbalance.
• To reduce overall system delay in MEC networks by optimizing task offloading and UAV trajectories.
• To enhance wireless communication coverage and service quality through dynamic UAV task offloading.

Main Methods:

• Developed a multi-UAV collaborative MEC network architecture integrating Non-Orthogonal Multiple Access (NOMA).
• Formulated a task offloading strategy optimization problem considering delay and energy consumption constraints.
• Designed a delay-optimized offloading strategy using the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm.

Main Results:

• The proposed TD3-based strategy significantly reduces overall system delay compared to traditional methods.
• Achieved delay reductions ranging from 9.8% to 20.2% across various scenarios (task volume, device numbers, UAV speed/time, computing capacities).
• Demonstrated effective load balancing and reduced queuing delays through dynamic inter-UAV task offloading.

Conclusions:

• The proposed multi-UAV collaborative MEC network with NOMA and TD3-based optimization is effective in minimizing system delay.
• Dynamic task offloading and optimized UAV trajectories are crucial for improving MEC performance.
• The solution offers a robust approach to address challenges in high-demand edge computing environments.