Energy Efficient Lossy Data Aggregation in Asynchronous Sensor Networks

In wireless sensor networks, most of existing data aggregation scheduling methods try to aggregate the data from all the nodes at all time-instances. It is neither energy efficient nor practical because of the link unreliability and spatio-temporal data correlation. This paper proposes a lossy data aggregation scheme to allow estimated aggregation at the root by selectively letting some nodes sample at some time slots. Firstly, all nodes sample data synchronously and the error between the real value and estimated one is guaranteed to being bounded respectively with and without the link unreliability. And the error bound is analyzed when the confidence is given a priori. Then we also design an algorithm to assign the confidence level among the parents such that each parent can calculate the minimum number of needed leaves based on the assigned confidence level. Secondly, all nodes sample data asynchronously, under which we analyze the probability that the error could be bounded under a given confidence level. Then a new algorithm is designed to implement data aggregation under a synchronization. We also present the experiment based on a real test-bed to evaluate our schemes.