Abstract: In delay tolerant networks, to deal with the problems that the topology of the network changes dramatically and the disconnections between nodes are prevalent, researchers propose the store-carry-forward protocols: nodes store the messages in buffers and may carry them for a long time until they encounter the proper next hops or the destination nodes. Because of the limited buffer capacity of nodes, this way of message transmission is bound to bring buffer overflows and then lead to network congestion. A congestion control strategy based on the game of life is proposed for delay tolerant networks in this paper and it is applied to classic Epidemic routing protocol. This strategy determines the specific operations of a message stored in the local buffer of node according to the proportion of the holders of this message in all the nodes’ neighbors. Furthermore, the policies of message queuing and dropping are designed. The utility value of a certain message is calculated based on the influence of delivering or dropping this message on the delivery ratio of the whole network, and the messages stored in the buffers are queued by the utility value. The messages with large utility values have high priority to send and the messages with small utility values are dropped. Experiments under movement model (which is the build-in model in the ONE) and real trajectories are carried out in the ONE. Simulation results show that the congestion control strategy based on the game of life significantly improves the delivery ratio compared with other congestion control strategies, while the delivery latency, packet loss rate and overhead ratio are reduced at the same time.