Research On Message Delivery And Buffer Management Based On Link Prediction In Opportunistic Networks

Opportunistic Networks is a new type of network architecture developed from Mobile Ad-hoc Network(MANET).It can use the opportunities brought by node mobility to implement hop-by-hop forwarding of data under split network conditions,and finally deliver them to destination node.Compared with traditional MANET,opportunistic networks do not require full network connectivity,and are more suitable for wide usage in practical self-organizing network application scenarios that lack basic communications equipment.However,because of its intermittent connectivity characteristics,it is necessary to use relay nodes to deliver messages in a “store-carry-forward” communication mode,and nodes can only maintain partial network topology information in distributed opportunistic networks,which resulting in low delivery success rate and high delivery delay.At the same time,limited node buffer,random message distribution,and multi-copy message delivery policies lead to severe buffer overflows and local network congestion,which greatly limit network performance.To solve these problems,the thesis studies the link prediction method in opportunistic networks,and applies it to the optimization design of message delivery strategy and buffer management mechanism.The main research work of the thesis is as follows:1)Designing a Stacking-based link prediction method named SMLP.The effective feature space is extracted fully considering the sparsity,dynamic and temporality in opportunistic networks,the rule of network topology evolution is modeled using the Stacking algorithm,and a sample unbalanced strategy UE for link sparsity is proposed,which predict the link probability of the next time slice efficiently and accurately.On the trace dataset,it is verified that the SMLP method performs better than others.2)Designing a message delivery strategy based on link prediction named LPMD.The SMLP method is used to predict the link probability and define the node's relaying capability for the message,and the delivery effect is defined to select the relay node combining the node's buffer remaining rate.According to the node's delivery utility,a threshold is set to limit message delivery and the number of messages copies in the node is dynamically allocated.3)Designing a buffer management mechanism based message priority named BPBM.Considering the message's time-to-live remaining rate,the proportion of message copies and the node's relaying capability for the message,the message priority is defined,then the buffer message scheduling mechanism and replacement mechanism is proposed.At the same time,a lightweight message redundancy copies deletion mechanism is proposed,which can promptly delete the messages that have been successfully delivered to the destination node,and thus alleviate the waste of network resources and node buffer.4)The LPMD strategy and BPBM mechanism are implemented on the ONE simulation platform,and simulation experiments are carried out by loading the trace dataset.Compared with other classic message delivery strategies and buffer management mechanisms,they has better network performance.