UAV Trajectory Control and Power Optimization for Low-Latency C-V2X Communications in a Federated Learning...

Area of Science:

• Wireless Communications
• 6G Vehicular Networks
• Artificial Intelligence in Communications

Background:

• 6G vehicular networks face challenges with line-of-sight (LoS) blockages, shadowed fading, and high data loads from applications like autonomous driving.
• Unmanned Aerial Vehicle (UAV)-assisted communication is crucial but impacted by mobility, shadowing, and limited UAV resources (battery, coverage).
• Data-intensive, non-i.i.d., and heterogeneous sensor data increase processing latency and demand efficient resource management on UAVs.

Purpose of the Study:

• To investigate system performance and communication disruption in UAV-assisted 6G vehicular networks, considering Doppler effect in Orthogonal Time-Frequency Space (OTFS) channels.
• To propose a low-complexity federated learning (FL) approach for UAV trajectory prediction and vehicle selection to enhance Quality of Service (QoS).
• To minimize UAV weighted total energy consumption by jointly optimizing transmission window, transmit power, and UAV trajectory.

Main Methods:

• Developed a federated learning algorithm for UAV trajectory prediction and vehicle selection, leveraging historical trajectory data.
• Investigated system performance under Doppler effect in OTFS modulated channels for UAV-C-V2X communication.
• Jointly optimized transmission window (Lw), transmit power, and UAV trajectory to minimize weighted total energy consumption.

Main Results:

• The OTFS-based system with federated learning achieved up to a 10% reduction in weighted total energy consumption by enabling local data processing on vehicles.
• The proposed federated learning algorithm demonstrated a 10-15% decrease in weighted total energy consumption compared to convex optimization, heuristic, and meta-heuristic methods.
• Federated learning effectively exploits related information from past trajectories for improved UAV operation and resource management.

Conclusions:

• Federated learning offers a viable solution for enhancing QoS in UAV-assisted 6G vehicular networks by optimizing UAV trajectory and vehicle selection.
• The proposed method significantly reduces UAV energy consumption while addressing challenges posed by mobility, shadowing, and data heterogeneity.
• This approach improves the efficiency and performance of data-intensive applications in future intelligent transportation systems.