The invention discloses an intelligent coverage
multicast routing method fusing an SDN (
Software Defined Network) and target-oriented hierarchical
reinforcement learning, which comprises the following steps: sensing the global state of an underlying
physical network in real time by utilizing an SDN controller, and decomposing a complex coverage
multicast tree construction task into two collaborative hierarchical subtasks by adopting a GO-HRL
algorithm; the upper-layer strategy network (target selection) is used as a high-layer
intelligent agent, and the lower-layer strategy network (path planning) is used as a low-layer
intelligent agent. And gradually constructing a complete coverage
multicast tree through an
iteration process of selecting a target at an upper layer and constructing a path at a lower layer. According to the method, global state
perception and dynamic self-adaption are realized, the problem of solving high dimension and path
coupling is solved, the learning efficiency and strategy stability are improved, and the comprehensive performance of the network is remarkably optimized.