Dynamic topology network intelligent routing method

The invention discloses a dynamic topology network intelligent routing method which is good in adaptability, supports complex routing constraints, and is high in efficiency. According to the technical scheme, the method comprises the following steps: a dynamic routing problem of a dynamic topology network is modeled into a reinforcement learning routing problem according to a dynamic routing process of the dynamic topology network, the network topology is used as a part of the state of the environment, a strategy function and a value function in a PPO reinforcement learning algorithm are approximated by using a graph neural network, the strategy function outputs weights of all links, and the network environment calculates a constrained minimum weight path based on the link weights, the routing agent performs interactive learning on generalization of different network topologies and a network environment to maximize an optimal routing strategy of network throughput, calculates a minimum weight path a meeting c