| Ad Hoc networks is a self-organizing, multi-hop networks composed of a group of node moving at any speed and with any motion model, which requires the routing protocol adapt to the change of network topology rapidly. The traditional AODV protocol uses the shortest path as their routing criteria, which results in routing load concentrating on just a few nodes. It leads to the discard or high transmission delay of the packet. So, it is important of the research on the load-balancing routing protocol.This thesis proposes a Dynamic Exponential Smoothing Model-Based Ad Hoc Network Traffic Prediction Load Balancing Protocol (DES-TPLBP). In the protocol, the traffic load according to the length of the interface queue of MAC layer is calculated and the dynamic average load of the network is defined as a threshold value. In the routing aegis stage, the nodes' load state in the future can be determined earlier by DES-TPLBP. Then the forthcoming congestion can be prevented by the route-discovery mechanism and the transfer reliability can be improved. In addition, the load value of the intermediate node is compared with the threshold value during the routing discovery stage, then the RREQ groups are selectively transfered on the basis of the comparison results to prevent the routing construction on re-load node. Meanwhile, the network performance is also improved by the delay response of destination node to some extent.Simulation was done on NS2-based platform for comparing the properties of DES-TPLBP, T-LBR and AODV protocol through three different angles. Simulation results proved that, in comparison with the traditional protocol, T-LBRP reduces the average end-to-end delay and the the normalized routing overhead by 32% and 18% respectively, while raising the delivery ratio by 8%. The DES-TPLBP protocol perform better than T-LBRP protocol, it can reduce the end-to-end delay by approximately 50% and the normalized routing overhead by 28%, while raising the delivery ratio by 10.4%. |
PDF Full Text Request