Abstract:

Routing maintains the stable operation of network infrastructure and supports the sustainable development of next-generation networks. Nowadays, the rapid growth of network traffic and the continuous changes in network services make traditional routing algorithms face severe challenges. In recent years, deep reinforcement learning has shown good results in solving complex continuous control problems. In order to solve a series of shortcomings of traditional routing algorithms, we combined the Deep Deterministic Policy Gradient (DDPG) algorithm with routing scenarios, and proposed a new dynamic routing algorithm based on deep reinforcement learning—DDPG4Net.

Key words: deep reinforcement learning, routing algorithm, network traffic engineering