Research On Active Queue Management And Wireless Sensor Networks

The Active Queue Management schemes are the link algorithms of the congestion control mechanism of IP networks. The AQM algorithms show TCP sender the impending congestion states of networks by dropping or marking packages actively. When receiving the congestion signal sent by AQM, TCP sender will take action for the impending congestion states in advance to avoid the network congestion. The clustering algorithms of Wireless Sensor Networks organize the sensor nodes into cluster structures, which can realize the hierarchical routing and topology control of WSNs. The motive of a clustering algorithm is to make all the sensor nodes consume their energy uniformly and die at the same time. Thus a good clustering algorithm can prolong the lifetime of WSNs.In the researches of the AQM field, the main methods used in improving RED algorithm are based on experience and emulational experiments. These simple research methods can't provide systematic and whole understanding of AQM algorithms. Some research results have been shown that it is difficulty to tune the parameters of RED algorithm and the small change of the parameters can affect the performance of RED algorithm heavily. At the same time, current AQM algorithms set parameters fixedly. They are sensitive to the change of network state parameters such as connection number, link capacity and round trip time. Improperly setting of the parameters of AQM algorithms will bring forth oscillation of queue length, depression of throughput and heavy jitter of queuing delay.In the researches of the clustering algorithms of WSNs, most clustering algorithms are homogeneous. That is to say they assume all the sensor nodes have the same initial energy. At the same time, these algorithms can't choose the cluster heads according to the residual energy. Under such homogeneous clustering algorithms, the low energy sensor nodes will die more quickly than high energy sensor nodes. This will lead to the blind area of the field sensed by WSNs and shorten the lifetime of WSNs. Especially in the heterogeneous environment. Current clustering algorithms can't utilize the total energy of the whole networks enough and appropriately arrange the chance of each node to be the cluster head. It is hardly for nodes to consume the energy uniformly in whether time or space standpoint.In this thesis, the reasons that cause the problems mentioned above have been analyzed. The optimal control theory and artificial neural networks control theory are introduced into the study of the AQM scheme. By the methods of theoretic analyses and emulational experiments, the themes of AQM schemes in congestion control and clustering algorithms in heterogeneous wireless sensor networks are explored and investigated. The adaptive AQM algorithms with good performance and the energy-efficient heterogeneous clustering algorithms have been proposed to overcome the problems existing in current schemes.The main results are as follows:1. The optimal control theory is used to optimize AQM scheme. Based on the operation of RED, a discrete mathematical model of AQM structure is proposed. Under certain target function, using the drop probability to be the control variable, the problem of optimizing the Active Queue Management can be converted into a dynamic programming problem and the optimal approach is proposed.2. The methods to design an adaptive AQM schemes are researched. To suppress the disturbance of unresponsive flows including HTTP and UDP connections, a novel AQM scheme is designed, which can minimize the output caused by disturbance based on the optimal control theory in frequency field. Moreover, a self-tuning structure used to complement the adaptive capability of the controller.3. Based on the artificial neural networks control theory, two new adaptive AQM algorithms are proposed to overcome the problems existing in PI and PID algorithms. A single neuron named ADALINE is used to adjust the parameters of the algorithms online. The new schemes have better performance and keep operation still simple compared with PI and PID algorithms.4. The definition of heterogeneous wireless sensor networks is given. The reasons that why need an energy-efficient algorithm for heterogeneous environment are analyzed. A new energy-efficient clustering algorithm for heterogeneous networks is proposed and evaluated. In the new clustering scheme, the cluster-heads are elected according to the initial energy and residual energy. It can ensure that each node expends its energy uniformly.