Federated Learning Semantic Communication in UAV Systems: PPO-Based Joint Trajectory and Resource Allocation...

Area of Science:

• Wireless Communication
• Artificial Intelligence
• Distributed Systems

Background:

• Semantic Communication (SC) reduces data transmission by sending semantic information (SI) instead of raw data, ideal for UAVs with Size, Weight, and Power (SWAP) constraints.
• Federated Learning (FL) offers a distributed training approach to offload computationally intensive tasks like semantic extraction (SE) from UAVs to edge users.

Purpose of the Study:

• To demonstrate the feasibility of integrating FL into SC systems for UAVs.
• To propose a novel Proximal Policy Optimization (PPO) based solution for service fairness in UAV-assisted SC.
• To jointly optimize UAV flight trajectory and bandwidth allocation for efficiency and fairness.

Main Methods:

• Developed a federated learning framework for distributed semantic extraction in UAV communication systems.
• Formulated a joint optimization problem for UAV trajectory and bandwidth allocation.
• Applied Proximal Policy Optimization (PPO) to ensure equitable Quality of Service (QoS).

Main Results:

• FL framework significantly reduces computational resource consumption on the UAV.
• PPO-based algorithm effectively minimizes energy consumption and task completion time.
• Equitable QoS is ensured across all edge users.

Conclusions:

• Federated learning is a viable approach to reduce computational burden in UAV-based semantic communication.
• The proposed PPO algorithm successfully balances efficiency and fairness in UAV-assisted SC systems.
• This work paves the way for more efficient and equitable UAV communication networks.