Efficient Data Collection Trees in Sensor Networks with Redundancy Removal

Abstract

In sensor networks, overlapping sensed areas result in the collection of partially redundant data. Interior sensor nodes along a data collection tree can remove redundancies and reduce the required bandwidth for the aggregate data. Assuming that the sensed information is proportional to the sensed area, we propose heuristic algorithms to build data gathering trees in order to reduce communication costs. The algorithms are completely distributed and are based on finding shortest hop paths and employing particular tie-breaking mechanisms. The mechanisms can be divided into two categories: those that use simple knowledge of the cardinality of a neighbor’s neighbor set and those that rely on knowledge of a distance–monotonic metric to a neighbor. We compare five heuristics based on communication cost and also on energy and processing costs. For smaller configurations, the performance of the heuristics are also compared to the optimal. With the proposed approaches, it is possible to build a cost efficient data gathering tree which takes advantage of the data redundancy among closely located sensors.