Multi-Objective Optimization in Air-to-Air Communication System Based on Multi-Agent Deep Reinforcement Learning.

Area of Science:

• * Focuses on aerial communication systems and mobile edge computing.
• * Integrates wireless power transfer for enhanced UAV endurance.

Background:

• * Unmanned aerial vehicle (UAV) systems face limitations in flight duration and computational capacity due to energy constraints.
• * Existing systems struggle to balance real-time data processing with sustained aerial operations.

Purpose of the Study:

• * To develop a full-duplex air-to-air communication system (A2ACS) for UAVs.
• * To reduce UAV computational latency and energy consumption while ensuring mission continuity.
• * To optimize system throughput, minimize low-power alerts, and manage energy transfer efficiently.

Main Methods:

• * Proposed a novel system combining mobile edge computing and wireless power transfer.
• * Developed a multi-objective optimization framework to balance system throughput, UAV power status, energy reception, and energy consumption.
• * Utilized a multi-agent deep deterministic policy gradient (MADDPG) algorithm for optimizing AEES location and energy transfer power.
• * Employed K-means clustering for fair association between air-edge energy servers (AEESs) and UAVs.

Main Results:

• * The proposed multi-objective DDPG (MODDPG) algorithm demonstrated superior performance compared to baseline methods.
• * The system effectively reduces computational latency and energy consumption for UAVs.
• * Achieved a balance between system throughput, UAV energy levels, and energy server efficiency.

Conclusions:

• * The integrated A2ACS model enhances UAV operational efficiency and endurance.
• * MADDPG-based optimization provides effective decision-making for dynamic aerial environments.
• * The K-means algorithm ensures equitable resource distribution, improving overall system fairness.