| This dissertation reports studies on the problems of route discovery, maintenance and load-balancing in Mobile Ad Hoc Networks (MANETs). A unified approach to solving these problems is the central theme. A series of approximations is used to model the performance of the network. This facilitates the development of a Statistically Multiplexed Adaptive Routing Technique (SMART) for MANETs, that adjusts to changes in the topology and the traffic pattern. SMART routes different proportions of traffic over multiple paths from a source to a destination. The criterion to determine the traffic proportions is the minimization of the mean end-to-end packet delay, in the approximate model. For this purpose, the control packets generated at every node continuously forage for new (alternative) routes.; Traffic in data networks is known to be bursty. Existing models, results, and approaches for generation of bursty traffic are surveyed and a new model is developed. This Nth order auto-regressive (AR) model (a) is simple and represents limited scale burstiness, (b) allows easy determination of parameters for any given degree of burstiness (the Hurst parameter), and (c) is computationally efficient to generate synthetic bursty traffic. Two different bursty traffic models (including our AR model) are used to demonstrate that SMART improves performance even under these adverse traffic conditions.; The performance of SMART is thoroughly examined with extensive simulation experiments using a highly mobile 30-node MANET under bursty traffic conditions. A few simplifications in route computations of SMART are developed and tested with simulations. Heterogeneous interconnections with long and short links are shown to increase scalability of MANETs. Results of simulation experiments with simplified SMART controlling data routing in such enhanced MANETs lead to the following general conclusions. In spite of several levels of approximations in modeling and computations, SMART deals concurrently with all of the following aspects: (a) Very bursty traffic, (b) high degree of mobility, and (c) restrictive conditions imposed by other network agencies. The performance of SMART, while dealing with all the above aspects, is very good in all the following categories: (a) Route discovery and maintenance, (b) average data packet delay, (c) average packet loss rate, and (d) average number of hops. |
