In wireless networks, throughput and capacity are not the only performance objectives. For example, in sensor networks, the goal is often not the throughput, but other metrics such as energy efficiency, network lifetime and event-reporting delay. Sensor networks also employ unique control functionalities such as sleep-wake scheduling, in-network data aggregation, and reliable broadcast. We have studied the problem of using anycast to improve the delay and network lifetime of low-duty-cycled event-driven wireless sensor networks. For a low-duty-cycled wireless sensor network (where nodes sleep most of the time and wake up occasionally), anycast can substantially reduce the event-reporting delay because a packet is forwarded to the earliest next-hop neighbor that wakes up. We have developed distributed algorithms that jointly control the sleep-wake pattern and the anycast routing policy to maximize the network lifetime subject to a constraint on the expected end-to-end event-reporting delay. Our optimal algorithm substantially outperforms existing heuristic solutions in the literature. In future work, we plan to pursue other types of problems in sensor networks, including in-network aggregation with lossy transmission links, and reliable broadcast that take into account network coding. We are interested in solutions with similar desirable features, such as loose-coupling, low-complexity, distributed implementation and provable efficiency.
Selected publications: