A contact prediction method for DTNs based on BP artificial neural network

Predicting the contact of the nodes in Delay Tolerant Networks (DTNs) helps to determine the next node of a data package and choose an appropriate transfer opportunity. The existing contact prediction methods mainly divide into two types, model-based methods and history-based methods. The model-based methods always need the location, velocity and direction of nodes which are difficult to obtain. So, this kind of methods can only suit one particular scenario, which don't have good adaptability. The history-based methods all consider the future contact has a linear correlation with the history contact, but in fact the future contact of nodes is also influenced by nodes' position, velocity, direction and other factors, in this way, the future contact shouldn't have a linear correlation with the history contact. In this paper, a contact prediction method for DTNs based on BP artificial neural network is proposed which uses BP neural network to predict the future contact of two nodes. This method includes two parts: discretization of time and design of BP neural network. The results show that this method can predict the future contact of two nodes more accurately than existing PROPHET.