Generally in ad hoc it is necessary for one mobile host to enlist the aid of others in forwarding a packet to its destination. In a fast changing ad hoc network, the routing tables will be out of data on a regular basis. Current routing algorithms are not adequate to tackle this problem. This study focuses on the use of topological information for routing in ad hoc network. The approach is to use topological information to train the Artificial Neural Network (ANN) for identifying the various subnets. Once the subnets are identified another ANN is used to elect the backbone node. This collection of backbone nodes forms the backbone network. A combination of AODV and DSDV protocols were used on the backbone network and on the local subnet for performance analysis. The AODV-DSDV protocols achieved a better delivery fraction and end-to-end delay than by using AODV alone. Implementation was carried out by using NS2. © 2006 Asian Network for Scientific Information.