This paper addresses the problem of how routers efficiently select the optimal routes in large-scale networks. For the first time, it proposes Aggregation Tree Routing Model (ATRM) for dynamic topology changes. Aggregation Tree is first constructed in this model based on various parameter characteristics of links between routers, which greatly narrows the searching space of the routing procedure within much smaller domains. When network state changes, protocol calls the increment solution presented in this paper, updating the both two kinds of state changes in dynamic networks including link cost changes with time and topology structure changes with time. This solution is composed of three sub-algorithms-Link Deletion Algorithm, Link Addition Algorithm and Link Change Algorithm, and merely modifies affected information of links without re-routing in the whole network, as well as satisfying QoS constrains. The work achieves a logarithmic reduction in communication complexity and the simulation demonstrates that this model obtains high performance in routing accuracy as expected.