Throughput optimization in wireless local networks with inter-AP interference via a joint-association control, rate control, and contention resolution

  • Authors:

  • Ka-Lok Hung;Brahim Bensaou

  • Affiliations:

  • -;-

  • Venue:
  • Ad Hoc Networks
  • Year:
  • 2014

Quantified Score

Hi-index 0.00

Visualization

Abstract

The dense deployment of wireless access points (APs) either in wireless local area networks (WLANs) or in wireless mesh networks facilitates greatly ubiquitous Internet access, however, due to the limited number of orthogonal frequency channels allotted to the IEEE802.11-based networks, this also induces the inevitable problem of inter-AP interference. In this paper, we study the problem of determining the optimal association in multi-cell or extended wireless networks in the presence of hidden terminals and inter-AP collisions. Unlike most previous work in this area, which deals with networks without inter-AP interference, we first reveal that association control alone is not sufficient to achieve fair throughput allocation and load balancing across APs, then advocate a solution based on the joint association control, rate control and fair contention resolution as a means to improving network performance. Based on this, we formulate a cross-layer association control, throughput optimization and contention resolution problem whose objective is to allocate downlink throughput according to the proportional fairness principle. As the problem turns out to be a non-convex mixed integer programming problem, known to be NP-hard, we relax it first into a continuous problem, then transform the resultant into a convex problem and finally propose a distributed algorithm to solve it. We then design a simple yet effective approximation algorithm to recover an optimal solution that fulfills the discrete integral association constraints. The algorithm yields the optimal association, the maximum achievable rate for each downlink flow as well as each AP's optimal average backoff time. Using these results as settings in a network, we can achieve the optimal operation point without any scheduling. Numerical experiments and simulation results show that our algorithm converges rapidly and works effectively.