Kalman Filtering in a Band width Constrained Sensor Network

The decentralized estimation problem ofdynamic stochastic process in a sensor network restrictedby communication bandwidth and energy is considered.Due to these constraints, only quantized messages of theoriginal information from local sensor are available. For thedynamic system composed by a state-vector model and aset of corresponding observation-vector models linked by anetwork, an adaptive quantization strategy and sequentialfiltering are used to design fusion algorithms. In terms ofdifferent forms of the transmitted information, two novelfusion filters based on the conventional Kalman filtering(KF) are presented by use of quantized measurements andinnovations respectively, abbreviated as KFQM and KFQI.In contrast, the latter has better estimation accuracy under the same bandwidth constraint condition. This is because that there is less information loss in the process ofquantizing innovations. Computer simulations show theeffectiveness of two novel filters.