A Distributed and Adaptive Signal Processing Approach to Reducing Energy Consumption in Sensor Networks Jim Chou, et al Univ. of Califonia at Berkeley Infocom ` /04/20 Presented by Hojin
2 Contents Introduction Distributed Compression Correlation Tracking Querying & Reporting Alg. Result Conclusion & Comment
3 Introduction Battery powered : energy depletion => network partition, data loss … Solution: energy aware-routing, efficient information processing … In this paper, they use inherent correlation(spatio- temporal) in sensor data
4 Distributed Compression(1/2) Each sensor can compress its data w/o knowing the other sensor’s data
5 Distributed Compression(2/2) 4-level tree code book(uncompressed data:4 bit) Using 2 bit X = 0.9(index 9) F(X) = 9 mod 4(2^2) = 1 => 01 Descend tree LSB first Assume(?) side info. Y = 0.8 Encoder: Decoder:
6 Correlation Tracking(1/3) Side information Y Use a Linear Predictive model Find the and that minimize the mean squared prediction error Assume and are pairwise jointly wide sense stationary
7 Correlation Tracking(2/3) Practically,
8 Correlation Tracking(3/3) If, no decoding error If, decoding error Chebyshev’s inequality
9 Querying & Reporting Alg.
10 Results(1/2) Light, temperature, humidity – each 18,000 samples Simulate the measurement of data by reading from a file, previously recorded from actual sensors 12 bit data Stat topology-1 data gathering node, 5 sensor nodes Prediction model:
11 Results(2/2) Zero decoding error => conservative in choosing i Spikes => chose aggressive weight factor Assume the energy used to transmit a bit is equivalent to the energy used to receive a bit
12 Conclusion & Comments Reduce energy consumption by using distributed compression and adaptive prediction Orthogonal approach to previous methods Overhead to data gathering node(computation & memory space)