Routing optimization in wireless mesh networks under uncertain traffic demands

Abstract

In this paper, we investigate the routing optimization problem in wireless mesh networks. While existing works usually assume static and known traffic demand, we emphasize that the actual traffic is time-varying and difficult to measure. In light of this, we alternatively pursue a stochastic optimization framework where the expected network utility is maximized. For multi-path routing scenario, we propose a stochastic programming approach which requires no priori knowledge on the probabilistic distribution of the traffic. For the single-path routing counterpart, we develop a learningbased algorithm which provably converges to the global optimum solution asymptotically.

Stochastic Traffic Engineering in Multihop Cognitive Wireless Mesh Networks

“…Moreover, our schemes enjoy a decentralized implementation, in contrast to centralized mechanisms in [24], [25] by utilizing the feedback signals and local information only. In our previous work of [26], we proposed a routing optimization scheme to combat with the randomness of instantaneous traffic in noncognitive wireless mesh networks. With respect to [26], this paper differs in the following ways: First, in the wireless mesh networks considered in [26], the capacity of each wireless link is assumed to be fixed, i.e., time invariant.…”

“…In our previous work of [26], we proposed a routing optimization scheme to combat with the randomness of instantaneous traffic in noncognitive wireless mesh networks. With respect to [26], this paper differs in the following ways: First, in the wireless mesh networks considered in [26], the capacity of each wireless link is assumed to be fixed, i.e., time invariant. However, in cognitive wireless mesh networks, due to the unpredictable appearance of primary users, the bandwidth of each wireless link is random.…”

“…However, in cognitive wireless mesh networks, due to the unpredictable appearance of primary users, the bandwidth of each wireless link is random. Second, in [26], the QoS requirement is not considered. Nevertheless, in this paper, we particularly address the QoS concern of each user, e.g., the expected accumulated delay on the paths cannot exceed a user-specific delay tolerance, as will be elaborated in Section 4.…”

“…Nevertheless, in this paper, we particularly address the QoS concern of each user, e.g., the expected accumulated delay on the paths cannot exceed a user-specific delay tolerance, as will be elaborated in Section 4. Third and most importantly, the analysis in [26] was based upon the assumption that all the users have convex utility functions. In this paper, we extend the techniques to address the scenarios with nonconvex utility functions.…”

“…However, in practice, the accurate measurement of probability distribution is a nontrivial task. In [26], we utilized a stochastic approximation-based approach to circumvent the difficulty of estimating the probability distribution. In the following, we will extend this technique and develop a tailored distributed algorithm to address the issues of time-varying link capacities as well as the user-specific QoS requirements, which are of particular interest in multihop cognitive wireless mesh networks.…”