Energy-efficient Data Compression In Wireless Sensor Networks

Energy is an important consideration in wireless sensor networks. The ubiquity of redundancies in the monitoring data inspires researchers to introduce compression technology for reducing data volume and saving communication energy costs. However, different application and network environments have a great impact on the energy efficiency of data compression. Maximizing the energy saving of compression algorithms becomes one of the most challenging issues for enhancing network lifetime. Therefore, in this dissertation, compression algorithms aimed at wireless sensor networks are studied systemically. Performance improvement of the compression algorithms and energy reduction of the networks are achieved by novel algorithm and architecture design.In many existing compression evaluations, compression ratio is still used and the complexity is often neglected. This leads to a subjective evaluation. A new criterion is proposed which focuses on the energy consumption of data compression, and a series of tunable compression algorithms are reevaluated. Based on the new criterion, a more objective evaluation is obtained, and the energy efficiency of each algorithm can be shown directly.Data compression in wireless sensor networks is not always beneficial to energy conservation. Consequently, a lightweight compression arbitration mechanism is designed which is suited for a single-hop network or a cluster topology. Based on the prediction of the balance point between loss and benefit, the mechanism improves compression algorithms by avoiding unnecessary energy losses. As an auxiliary algorithm, the mechanism makes compression more flexible, and the energy saving can be up to 40%.An adaptive compression arbitration mechanism is further proposed deal with the uncertainty in applications. On-line prediction modeling and adaptive mechanism are added into the new mechanism. Parameters of the prediction models can be modified dynamically. With acceptable costs in computation, the mechanism becomes more robust, and the highest energy saving is 30%.In order to save the total energy of the whole network, an energy optimization mechanism is proposed for a multi-hop network. With the feature extractions of network, data and compression algorithm, the total energy costs in data collection can be significantly reduced.Moreover, an energy-efficient architecture of sensor node is designed. It is based on a general-purpose processor (8051) and compression hardware accelerators. The sensor node is implemented by 1-poly 6-metal 0.18μm CMOS technology. The test results show that the new architecture with compression accelerator reduces over 98% energy consumption compared with the processor based solution.