Abstract:

A routing algorithm based on time series prediction for delay tolerant network is proposed. In this paper the weight moving average method which is one of time series prediction methods is improved, and the history connected information of nodes is used to predict the connection state in the future. Furthermore, the routing criterion is to ensure that messages reach the target node as much as possible soon. Any network connection information does not need, and the routing path can change followed with the change of network topology so that new routing table can be update immediately. The proposed algorithm can avoid the shortcoming of existing algorithm which needs amount of prior knowledge of routing information. So the proposed algorithm has strong applicability. The simulation experiment results have showed that the proposed algorithm compared with MED and MEED have higher delivery ratio and lower average time in different bandwidth and buffer size.