Joint optimization of UAV trajectory planning, video cache placement and transcoding in UAV-assisted 6G networks: a PPO-L based approach

In the context of integrated aerial-terrestrial networks, the use of unmanned aerial vehicles (UAVs) as aerial base stations (ABS) to provide additional edge computing and communication capabilities to ground users (GDs) is envisioned as a promising solution. Meanwhile, with the rapid development of 6G networks, there has been an explosive growth in GD demand for videos. Edge caching has been proposed to reduce redundant video transmission in the network and minimize GD waiting latency. In this paper, we investigate a video adaptive caching solution in UAV-assisted aerial-ground networks. Our objective is to minimize GD waiting latency by jointly optimizing caching decisions for both UAVs and base stations (BSs), as well as the trajectories of the UAVs. Due to the complexity of the proposed problem, which is difficult to solve by traditional optimisation methods, and the highly dynamic nature of video requests, we propose a reinforcement learning algorithm based on proximal policy optimisation and a long short-term memory neural network (PPO-L) to solve the problem. We use a real-world YouTube video request dataset as our simulation experiment dataset. Through a large number of simulation experiments and performance comparisons with other algorithms, we demonstrate the superiority of the proposed algorithm. The simulation results demonstrate that the proposed algorithm significantly improves performance in system experiments compared to other algorithms.