SDN-DRLTE Algorithm Based on DRL in Computer Network Traffic Control

Abstract

Affected by mobile Internet, big data and cloud computing, network traffic load is gradually increasing, and deep reinforcement learning algorithm has been widely used. To solve the uneven and congested computer network traffic, a software-defined network algorithm on the basis of deep reinforcement learning is designed, and a computer network traffic control technology is built. On the basis of traditional deep reinforcement learning algorithms, the optimal performance policy is obtained by combining Markov decision. Simultaneously, the Off Policy is introduced to establish a software-defined network traffic control model, ultimately designing a software-defined network algorithm on the basis of deep reinforcement learning. In contrast with other methods, the designed method had a higher average speed and significantly reduced latency. The average reward value of the algorithm was 12.2%, 18.6%, and 6.8% better than other algorithms, and the reward value increased linearly at 3000 iterations. This indicated that the designed algorithm achieved the expected goals in terms of computational efficiency and network scheduling control performance. The research findings were of great significance for computer network traffic control.