Robustness and stability of utility maximization algorithms for MANETs

We study different network utility maximization algorithms that optimize the medium access probabilities at the MAC layer jointly with the end-to-end source rates at the transport layer. The algorithms work in a multihop random access network in a distributed fashion to achieve different types of fairness for the users. Although the algorithms are proven to be mathematically stable, they never have been tested on dynamic network topologies with imperfect transmission links and asynchronous parameter updates. An object-oriented simulation model with modifications to ensure and increase the speed of convergence is proposed. We report on stability tests with different link failure scenarios, topology changes and session rerouting in case of synchronous and asynchronous message updates. The network behavior under different network parameters is discussed. The results show that the modified algorithm converges even under conditions of high link transmission failure rates and can tolerate asynchronous updates of contention probabilities.