Abstract: To solve the edge caching problem of cloud server and edge server in panorama video service, optimizing the edge caching mechanism to reduce the time delay for obtaining video resources, a method of generating cache strategy by using DQN as deep reinforcement learning algorithm is proposed. First, the problem is modeled as Markov decision process with the goal of total time saving. Then, DQN algorithm is used for training to obtain the best cache strategy in the iteration. Simulation shows that DQN algorithm has high convergence speed and the best performance. And when the constraints change, it can actively change the edge caching strategy to stably improve the performance of the algorithm.

Key words: edge computing, deep reinforcement learning, cache strategy, panorama video