Abstract: Location management is used in mobile communication networks to track mobile terminals (MTs). It has two basic operations: location update and paging. Under existing location management schemes, an entire location area consisting of a lot of cells is paged concurrently to locate an MT, which consumes dramatically networks’ resources, since the MT only resides one of the paged cells. Based on reinforcement learning, multi-stage intelligent paging schemes are proposed, which learns MTs’ location information obtained in previous paging processes and the cells included in each paging stage are dynamically adjusted, so that paging cost is reduced. Simulation results illustrate that the paging cost can be reduced considerably by proposed multi-stage paging schemes.