Onboard Deep Deterministic Policy Gradients for Online Flight Resource Allocation of UAVs

• Autor(a) principal: Li, Kai
• Data de Publicação: 2020
• Outros Autores: Emami, Yousef, Ni, Wei, Tovar, Eduardo, Han, Zhu
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10400.22/16331
• topic: Unmanned aerial vehicles; Flight control; Data collection; Deep reinforcement learning

In Unmanned Aerial Vehicle (UAV) enabled data collection, scheduling data transmissions of the ground nodes while controlling flight of the UAV, e.g., heading and velocity, is critical to reduce the data packet loss resulting from buffer overflows and channel fading. In this letter, a new online flight resource allocation scheme based on deep deterministic policy gradients (DDPG-FRAS) is studied to jointly optimize the flight control of the UAV and data collection scheduling along the trajectory in real time, thereby asymptotically minimizing the packet loss of the ground sensor networks. Numerical results confirm that the proposed DDPG-FRAS can gradually converge, while enlarging the buffer size can reduce the packet loss by 47.9%.