Date on Master's Thesis/Doctoral Dissertation


Document Type

Doctoral Dissertation

Degree Name

Ph. D.


Computer Engineering and Computer Science

Committee Chair

Kumar, Anup

Author's Keywords

Ad-hoc network; EM-routing; Stitching algorithm; MLE; Expectation-maximization


Electric network topology; Mobile communication systems; Wireless communication systems; Mobile computing


The memorability of mobile ad-hoc network (MANET) is the precondition of its management, performance optimization and network resources re-allocations. The traditional network interior measurement technique performs measurement on the nodes or links directly, and obtains the node or link performance through analyzing the measurement sample, which usually is used in the wired networks measurement based on the solid infrastructure. However, MANET is an infrastructure-free, multihop, and self-organized temporary network, comprised of a group of mobile nodes with wireless communication devices. Not only does its topology structure vary with time, but also the communication protocol used in its network layer or data link layer is diverse and non-standard. Specially, with the limitation of node energy and wireless bandwidth, the traditional interior network measurement technique is not suited for the measurement requirement of MANET. In order to solve the problem of interior links performance (such as packet loss rate and delay) measurement in MANET, this dissertation has adopted an external measurement based on network tomography (NT). Being a new measurement technology, NT collects the sample of path performance based on end-to-end measurement to infer the probability distribution of the network logical links performance parameters by using mathematical statistics theory, which neither need any cooperation from internal network, nor dependence from communication protocols, and has the merit of being deployed exibly. Thus from our literature review it can be concluded that Network Tomography technique is adaptable for ad-hoc network measurement. We have the following contribution in the eld of ad-hoc network performance: PLE Algorithm: We developed the PLE algorithm based on EM model, which statistically infer the link performance. Stitching Algorithm: Stitching algorithm is based on the isomorphic properties of a directed graph. The proposed algorithm concatenates the links, which are common over various steady state period and carry forward the ones, which are not. Hence in the process it gives the network performance analysis of the entire network over the observation period. EM routing: EM routing is based on the statistical inference calculated by our PLE algorithm. EM routing provides multiple performance metric such as link delay and hops of all the possible path in various time period in a wireless mesh network.