Research On Cluster-based Networking Technology In Mobile Ad Hoc Networks

With the recent developments in wireless communication technologies and evolutions in low cost wireless devices such as mobile phones, PDAs, and laptops etc., the human society is entering a ubiquitous communicating era. Among the various possible solutions, cluster-based Mobile Ad Hoc Networks (MANETs) is a proven one for deploying a network without any fixed infrastructure in a mobile scenario. Compared with the traditional flat-structured counterpart, the cluster-based network structure can dramatically improve the channel efficiency, reduce the routing complexity, and enhance the network stability. However, to initialize or maintain a cluster structure in a dynami-cally changing scenario often requires explicit message exchanges between mobile node pairs. When the underlying network topology changes quickly and involves many mo-bile nodes, the clustering-related information exchange increases drastically, which may even block the data transmissions. Therefore, the design of efficient, reliable cluster-ing scheme, and the accurate performance analysis for MANETs are vital for the further development of cluster-based networking technologies.In this dissertation, a Mobility Prediction Based Clustering (MPBC) scheme is firstly proposed to alleviate the effect of random mobility on network topology. Compared with other existing mobility-aware clustering schemes, MPBC adopts a Doppler shift based relative mobility prediction scheme, which works independently from the external system and without extra data interactions, to obtain the required mobility information. Besides its function for initial cluster head selection, the mobility-related information are farther used in the cluster maintaining stage of MPBC. By theoretical analysis and simulation experiments, MPBC outperforms other clustering schemes significantly in association with link status change rate, logic link lifetime, cluster head lifetime, and other stability performance metrics.Due to that most of the control message exchanges for clustering are carried out by broadcast, the broadcast transmission’s reliability plays a key role in clustering schemes’ implementation. To accurately illustrate the broadcast progress, this work models the ran-dom backoff counter’s state transition and the changing pattern of the number of packets waiting in a node’s transmission queue as two Markov chains, respectively. Afterwards, by using the transmission probability derived from the two new Markov models, three different reliability metrics, namely Distance-based Packet Loss Rate (d-PLR), Average Packet Loss Rate (APLR), and Average Queueing Delay (AQD) are analyzed. The deriva-tions for the closed-form of these metrics combine with the features of Poisson Point Process, and take into account both the impact of concurrent transmissions caused by identical backoff interval selection and that of hidden nodes caused by limited sensing range on an observed transmitter. Moreover, a short discussion for the effect of fading channel on broadcast reliability is also given after the part of simulation evaluations.Finally, this work studies the capacity bounds of the clustered MANETs. By making use of the concepts and features of Exclusive Region and Protocol Model, the two theo-retical upper bounds of network capacity induced by clustering and orphan nodes in the network are achieved first. After calculating the critical network area, two constructive lower bounds for network capacity are also obtained. The main result is that the clus-tered network capacity switches behavior at a critical network size, which depends on the cluster size and the density of clusters within the network. Moreover, the clustering parameters of the network allow us to define the meaning of a large network by certain bounds.