In order to reduce the hoop count and increase search efficiency,the notion of pointer table is introduced into CAN.In a scale of 2L identifier space,coordinates in each dimension are divided with binary search.The nodes corresponding to the divided coordinates compose the next hoop set as a pointer table,which cuts down the search scope from the whole CAN to a local CAN area.Simulation results show that the distribution of nodes’coordinates generated by improved CAN search algorithm is more uniform than that of the original CAN model.In the scale of 26 CAN and the scale of 27 CAN,90% and 70% searches cut their hoop figures,and the rates of decrease are 53.2% and 31.5% respectively.The distribution rates of search length reduction in the scales of 25,25 and 27 CANs are given after the sample space is extended.The experiment demonstrates that the improved algorithm based on CAN has less search hoop count than the original one.