Paper detail

Harvest: A Reliable and Energy Efficient Bulk Data Collection Service for Large Scale Wireless Sensor Networks

We present a bulk data collection service, Harvest, for energy constrained wireless sensor nodes. To increase spatial reuse and thereby decrease latency, Harvest performs concurrent, pipelined exfiltration from multiple nodes to a base station. To this end, it uses a distance-k coloring of the nodes, notably with a constant number of colors, which yields a TDMA schedule whereby nodes can communicate concurrently with low packet losses due to collision. This coloring is based on a randomized CSMA approach which does not exploit location knowledge. Given a bounded degree of the network, each node waits only O$(1)$ time to obtain a unique color among its distance-k neighbors, in contrast to the traditional deterministic distributed distance-k vertex coloring wherein each node waits O$(Δ^{2})$ time to obtain a color. Harvest offers the option of limiting memory use to only a small constant number of bytes or of improving latency with increased memory use; it can be used with or without additional mechanisms for reliability of message forwarding. We experimentally evaluate the performance of Harvest using 51 motes in the Kansei testbed. We also provide theoretical as well as TOSSIM-based comparison of Harvest with Straw, an extant data collection service implemented for TinyOS platforms that use one-node at a time exfiltration. For networks with more than 3-hops, Harvest reduces the latency by at least 33% as compared to that of Straw.