How to guarantee the transport efficiency of the network and how to improve the network capacity are the main subject of the study presently. We investigate the gravity of the nodes to the transfer of data packets, and propose a routing method based on gravity constraint. In order to characterize the efficiency of the method, we introduce an order parameter H to measure the throughput of the network by a critical value of phase transition from free state to jammed state, and use the maximum travel time 〈Tmax〉 and the average travel time 〈Tavg〉 to test the transmission efficiency of the network. We simulate the network capacity under three different gravity constraints. Simulation results show that when only considering the path with shortest length, the network capacity is very small and the distribution of flow is extremely uneven; when only considering minimum waiting time, the excessive circuitous transfer of data packets occurs and most of the nodes will be in congestion state; when considering the gravity of path length and waiting time simultaneously and choosing a node with reasonable gravity, the network capacity will be improved greatly and the congestion level will be relieved to some extent.