Data Compression in Multi-hop Large-scale Wireless Sensor Networks

Data collection from a multi-hop large-scale outdoor WSN deployment for environmental monitoring is full of challenges due to the severe resource constraints on small battery-operated motes (e.g., bandwidth, memory, power, and computing capacity) and the highly dynamic wireless link conditions in an outdoor communication environment. We present a compressed sensing approach which can recover the sensing data at the sink with good accuracy when very few packets are collected, thus leading to a significant reduction of the network traffic and an extension of the WSN lifetime. Interplaying with the dynamic WSN routing topology, the proposed approach is efficient and simple to implement on the resource-constrained motes without mote’s storing of a part of random measurement matrix, as opposed to other existing compressed sensing based schemes. We provide a systematic method via machine learning to find a suitable representation basis, for the given WSN deployment and data field, which is both sparse and incoherent with the measurement matrix in the compressed sensing. We validate our approach and evaluate its performance using our real-world multi-hop WSN testbed deployment in situ in collecting the humidity and soil moisture data. The results show that our approach significantly outperforms the k-nearest neighbors weighted regression regarding the data recovery accuracy for the entire WSN observation field under drastically reduced communication costs. For some WSN scenarios, CS may not be applicable. Therefore we also design a generalized predictive coding framework for unified lossless and lossy data compression. In addition, we devise a novel algorithm for lossless compression to significantly improve data compression performance for various data collections and applications in WSNs. Rigorous simulations show our proposed framework and compression algorithm outperform several recent popular compression algorithms for wireless sensor networks such as LEC, S-LZW and LTC using various real-world sensor data sets, demonstrating the merit of the proposed framework for unified temporal lossless and lossy data compression in WSNs