No abstract

We develop a rate adaptation scheme for 802.11 networks termed SRC that is based on a novel combination of sequential hypothesis testing and short-term loss ratios. The underlying design of SRC represents a fundamental departure from previous rate adjustment strategies that employ fixedsize observation windows and seemingly-intuitive rate increase guidelines. Through extensive simulation results, SRC is shown to be robust to dynamic characteristics of the wireless channel and to significantly increase network throughput over existing mainstream rate adaptation solutions. SRC does not require any type of specialized hardware or modifications to the existing 802.11 standard.

Abstract-The complementary characteristics of 3G cellular networks (such as UMTS) and IEEE's 802.11 wireless LANs (WLANs) have stimulated intensive research efforts to integrate UMTS and WLAN networks. In such integrated networks, a mobile station is able to switch between UMTS and WLAN networks based on various network parameters such as network throughput, packet delay and other network or user quality criteria. In this paper, we develop a utility-based access selection algorithm targeted to achieve load balancing between UMTS and WLAN networks. The performance of our algorithm is evaluated using OPNET [1] with dual-mode mobile stations.

No abstract

Ad hoc networks employ a decentralized unstructured networking model that relies on node cooperation for key network functionalities such as routing and medium access. In this paper, we develop a model based on the Sequential Probability Ratio Test to characterize how nodes can differentiate between routes that include misbehaving nodes (infected routes) and routes that do not. An advantage of the model is that the number of observations required to evaluate a route need not be determined in advance, which suits well the dynamic nature of ad hoc networks. We then outline a centralized and a localized approach to detect misbehaving nodes on infected routes identified by the model. Our evaluation shows that the localized approach is not only the better architectural choice for ad hoc networks but also results in a more accurate exposure of misbehaving nodes while incurring low false positives and low false negatives.