Abstract: In order to overcome the problem that there are too many message copies and large data transmission delay in the delay tolerant network, the author optimizes the Prophet routing protocol based on historical prediction, and propose a routing protocol based on density clustering. The algorithm adopts cluster analysis theory and birth and death process theory, the density cluster is constructed and maintained accurately, so that the message copies in the network can be controlled in real time. On this basis, the random linear network coding strategy based on Q-learning is proposed. The value function estimation method in the enhanced learning domain is adopted to obtain the linear independent coding packets efficiently through the intermediate nodes, so as to improve the coding efficiency of the network. Simulations show that this algorithm can obtain a higher message delivery rate in comparison with epidemic and Prophet routing algorithms, the data transmission delay can be well controlled under the condition of sufficient buffer thereafter. The algorithm has strong dynamic adaptability to the delay tolerant network.

Key words: delay tolerant network, Prophet routing protocol, density clustering, the random linear network coding