We present an effective distributed power scheduling algorithm for fixed, low bandwidth, many to one data collection sensor network applications. DuraNet reduces energy consumption by avoiding collision and overhearing while having nodes sleep most of the time. Because that it is hard to achieve global time synchronization while nodes are sleeping, DuraNet avoids the traditional approach where global hard bound time slots are assigned to nodes for communication. Instead, there is no notion of global time in DuraNet. Nodes allocate schedules for each link based on contention in a schedule formation phase, then re-use these schedules for a long period of time. By doing so, DuraNet is able to adapt to density, while providing significant energy savings. Given a route tree topology, DuraNet is also able to avoid congestion due to buffer queue overflow by careful scheduling.
We compare our algorithm to low power listening and classic CSMA through extensive simulation in TOSSIM. We find that DuraNet achieves significant energy reduction over classic CSMA and low power listening, while providing similar end-to-end reliability.