Distributed Compressed Sensing for Sensor Networks with Packet Erasures

Paper in proceeding, 2014

We study two approaches to distributed compressed sensing for in-network data compression and signal reconstruc- tion at a sink. Communication to the sink is considered to be bandwidth-constrained due to the large number of devices. By using distributed compressed sensing for compression of the data in the network, the communication cost (bandwidth usage) to the sink can be decreased at the expense of delay induced by the local communication. We investigate the relation between cost and delay given a certain reconstruction performance requirement when using basis pursuit denoising for reconstruction. Moreover, we analyze and compare the performance degradation due to erased packets sent to the sink.