Congestion Control, Ad Hoc Networks

Mobile Ad Hoc networks, which can be set up rapidly, are centerless and self-organized. Because of these special characteristics, Ad Hoc networks can be widely used in battlefield or any other emergency conditions. Mobile Ad Hoc networks are also characterized by dynamic topology due to node mobility, limited channel bandwidth and limited battery power of nodes, which make it difficult to provide high QoS for data transmission in Ad Hoc networks.In order to improve the performance of data transmission in Ad Hoc networks, the characteristics of the data flow have been introduced into the designing of routing protocol by researchers. With the prevalence of multimedia application, it is necessary to transmit audio and video data flows which are self-similar in Ad Hoc networks. The self-similar data flow makes Ad Hoc networks congested. And the congestion leads to the overflow of MAC layer interface queue, the aggravation of collisions in MAC layer, and even the disconnection of paths due to the lost of routing control packets. In this thesis, the self-similar data flow is generated through the superposition of many Pareto ON/OFF data source models, and it is used to simulate the performance of AODV routing protocol under self-similar data flow circumstance in Ad Hoc networks. The simulation results show that the higher the degree of self-similarity is, the poorer the performances in average end to end delay, packets delivery fraction and routing overheads of the routing protocol are.Based on the impact of the self-similarity of data flow on the routing protocol, a new load balanced multipath routing protocol AOMDV-LB derived from AOMDV is proposed in this thesis. Many paths from source node to destination node are set up in AOMDV routing protocol. When the transmission path is broken off, the source node may switch to the best path left in the routing table. The packets transmission delay and the routing overheads decrease. The disconnection of paths due to the change of the topology is well adapted. But AOMDV routing protocol just uses the setup time of the path as its path selection policy and doesn't take the load of the path into account. At the same time, the load-aware policies of existing load-aware routing protocol are summarized. It is found that some of the existing load-aware policies can't reflect the congestion degree of the path completely, and others are so complex that the processing time in the node and the data transmission delay increase, the throughput of the network decreases. The method of setting up multiple paths in AOMDV routing protocol is introduced into AOMDV-LB. Meanwhile a new load-aware policy is used to compute the priority for every path. AOMDV-LB is designed through crosslayer ideas. The joint model between the average interface queue length in MAC layer of the intermediate nodes in the path and the round-trip time of the routing control packets is set up to compute the priority of each path. The factor a is used to balance the effect of the average interface queue length in MAC layer of the intermediate nodes of the path and the round-trip time of the routing control packets on the computation of priorities. The larger the value ofαis, the greater effect the round-trip time of the routing control packets has on the computation of priorities.The performances of AOMDV-LB routing protocol are validated using NS2 simulator in this thesis. The simulation results are as follows:1. When varying mobility, AOMDV-LB increases packets delivery fraction by1.7%, reduces packets transmission delay by 15.5% and normalized routing overheads 7.3% compared with AOMDV, and AOMDV-LB increases packets delivery fraction by 4.7%, reduces packets transmission delay by 23.2% and normalized routing overheads by 20.3% compared with AODV.2. When varying network load, AOMDV-LB increases packets delivery fraction by 0.9%, reduces packets transmission delay by 14.7% and normalized routing overheads by 1.9% compared with AOMDV, and AOMDV-LB increases packets delivery fraction by 3.4%, reduces packets transmission delay by 11.7% and normalized routing overheads by 10.5% compared with AODV.The balance factor in AOMDV-LB is also optimized in this thesis. The value ofαis fixed through experiments. So AOMDV-LB routing protocol has better performances in packets delivery fraction, average end to end delay and normalized routing overheads.