Shared Backbone PPO Algorithm Enhances Multi-UAV Communication Coverage
Researchers propose a Shared Backbone Proximal Policy Optimization (PPO) algorithm for multi-UAV communication coverage tasks. The algorithm shares the base module between Actor and Critic networks, enabling efficient training and improved performance over standard PPO. A graph information aggregation module is integrated to handle communication conditions among agents, fostering higher cooperation in the swarm. The method is tested in a connectivity-preserving multi-UAV swarm scenario, demonstrating superior results.
Key facts
- Shared Backbone PPO algorithm proposed for multi-UAV communication coverage.
- Base module shared between Actor and Critic networks for efficient training.
- Algorithm compared with standard PPO, achieving superior performance.
- Graph information aggregation module incorporated for agent communication.
- Trained agent swarm exhibits higher level of cooperation.
- Task involves connectivity-preserving multi-UAV swarm communication coverage.
Entities
Institutions
- arXiv