PAST RESEARCH PROJECTS
- Runtime Tuning of Contention Parameters
In a contention based wireless network (e.g. IEEE 802.11), the amount of bandwidth that a node can get depends on the activities of its neighboring nodes. By assigning different contention parameters (e.g. contention-window size, inter-frame space etc) to different nodes (or flows), service differentiation and high channel utilization can be achieved. I have been analyzing the relationship between these tunable parameters and the resultant bandwidth allocation and have designed a distributed algorithm to dynamically adjust these parameters to achieve the desired fair bandwidth allocation to best effort flows and QoS guarantees to realtime flows while maintaining high channel utilization in a dynamic environment.
- Bandwidth Prediction in Multi-Priority Wireless Networks
Predicting the achievable bandwidth of a new flow and its impact on existing flows are both crucial for bandwidth-aware services, such as load balancing, admission control, routing, etc. In a multi-priority contention-based wireless network (e.g. IEEE 802.11e), such prediction must be based on the contention behavior of flows, which is related to the priorities and traffic types of the flows. I have been analyzing the relationship between the contention behavior of flows and their bandwidth allocations and have designed a novel channel model that can be used to accurately predict the achievable bandwidth and network impact of a new flow.
- Interference Aware Protocol
In a wireless network, interference may happen among nodes that do not know each other directly, since they may be outside each others communication range but inside each others interference range. When performing QoS routing or admission control in such a network, the effect of interference must be considered. I have been studying the effect of interference on QoS routing and admission control in ad hoc networks and have designed interference aware protocols to support QoS in such networks.
- Channel State Adaptation of TCP
In a wireless LAN, the channel state experienced by a flow often varies dramatically due to fading. Instead of using scheduling algorithms to exploit this characteristic, which has been extensively studied by other researchers, my research focused on the possibilities of adapting TCP according to the channel state to achieve the same goal of the channel-state-dependent scheduling algorithms.
- Communication Resource Discovery
Currently many mobile devices are equipped with multiple communication interfaces. (e.g. a laptop may be able to communicate through infrared, IEEE 802.11, Bluetooth, and wired network.) Several protocols have been proposed to exploit the multi-interface ability of mobile devices to improve the QoS. However, as a mobile device roams around, it has to periodically wake up each interface to check the availability of communication resources in its environment. This is very expensive in terms of energy consumption. We proposed to build a map of available communication resources. The mobile device can first download this map into its memory and then decide which interfaces to wake up based on its location in the map.