A Non-Convex Distributed Optimization Framework and its Application to Wireless Ad-hoc Networks

The continuously increasing demand for resources in modern, both wired and wireless, communication networks urges for more efficient resource allocation. Such an allocation of resources to network users can be formulated as an optimization problem. Traditional resource allocation protocols, such as TCP, operate inefficiently in cases that there is competition for resources by multimedia applications and some, or possibly all, links in the network are wireless. In this paper, the performance degradation of TCP in modern networks is quantified to highlight the necessity for a novel optimization-based resource allocation protocol. To this direction, a new optimization framework is presented that can provide the theoretical foundations of such a protocol by proving a sufficient, and in some cases also necessary, condition for distributed solution of non-convex problems. The wide applicability of this general framework is illustrated by considering a resource allocation formulation in TDMA/CDMA ad-hoc networks. The convergence properties to the optimal solution are first identified and a distributed algorithm is proposed. Moreover, a novel heuristic is developed to approximate the optimal solution when the condition does not hold and resolve network oscillations. Finally, the performance of the proposed methodology is evaluated and compared against other approaches in literature by simulation.