Pkf-st: A communication cost reduction scheme using spatial and temporal correlation for wireless sensor networks

One of the most energy-intensive processes in wireless sensor networks (WSNs) is radio communication, which can be minimized with data compression techniques by using the inherent existence of spatial and temporal correlations in the physical phenomena. Exploiting the two correlation types with low algorithmic overhead in a distributed scenario is challenging. This work proposes PKF-ST, a technique that uses a predictor combined with Kalman filter (KF) to reduce the transmission rate for cluster-based WSNs. Each leaf node uses a reduced-order state space model to independently compress its own data based on temporal correlation. To improve the reconstruction quality, the cluster head uses a KF with the full-order model. Without any intra-communication, the energy cost of each node is further reduced with the help of the spatial correlation. Compared to traditional temporal compression algorithms, PKF-ST maximizes the utilization of temporal correlation; compared with the techniques using spatial correlation, PKF-ST works independently of the networks size and without any coordinator. The simulation results with both artificial signals and real temperature values demonstrate the efficiency of PKF-ST. Compared with a previous technique using spatial-temporal compression, it reduces the reconstruction error by 75.8 %.