Randomized trial demonstrates improved communication reliability in drone swarms, suggesting enhanced efficiency in challenging environments.
Key Points
This research aims to enhance communication reliability and efficiency in UAV swarms in complex environments.
Proposed a Curriculum Learning-assisted Multi-Agent Proximal Policy Optimization (CL-MAPPO) algorithm.
Utilized a Centralized Training with Decentralized Execution (CTDE) architecture for spectrum cooperation.
Developed a three-stage progressive curriculum learning mechanism focused on collision avoidance, load balancing, and dynamic anti-jamming.
The CL-MAPPO outperformed baseline models including Carrier Sense Multiple Access (CSMA) and random frequency hopping in throughput and collision rates.
Significant improvements in convergence speed compared to Multi-Agent Deep Deterministic Policy Gradient (MADDPG).
Demonstrated effectiveness in scenarios with dynamic sweep jamming and multi-drone communication.