Joint Optimization of Trajectory, Caching and Task Offloading for Multi-Tier UAV MEC Networks

With the explosive growth of compute-intensive and time-sensitive applications, edge service caching has been recognized as an effective solution. In this paper, we study a two-tier unmanned Aerial vehicle(UAV) system supporting service caching, where the lower-tier UAV acts as an airborne base station and the upper-tier UAV acts as a small cloud server. In order to improve quality of service for users and increase UAV endurance time, we propose a two time-scale decision framework to jointly optimize system latency and energy consumption. The framework consists of a short time scale and a long time scale, where the lower-tier UAV cache placement and ground user task offloading decisions are updated in the short time scale (i.e., each system time slot t). In the long time scale
$\mathbf{T}$
, the cache placement and flight trajectory decisions of the upper-tier UAV are updated. We then propose a two-tier decision-making algorithm based on deep deterministic policy gradient (DDPG) to solve the problem, and use the long and short-term memory neural networks(LSTM) algorithm to predict the service content information requested by the ground user. Simulation experimental results show that the proposed algorithm can significantly reduce the system latency while maintaining low energy consumption compared to other schemes.