Iterative Model Identification and Tracking with Distributed Sensors

In this paper, we propose a new iterative model identification and tracking technique for distributed sensor systems using a factor graph (FG). The idea is initiated from a position identification technique we proposed [4], however, this paper aims to provide an algorithm which is applicable to more generic identification and tracking purposes. With the proposed technique, each sensor performs signal processing for compression of the measurement data and sends the compressed data to the fusion center. The marginal probability of the compressed sensing results are calculated over the FG at the fusion center. At the final stage, a maximum a posteriori probability (MAP) estimate of the model can be obtained during the tracking phase through the FG. The MAP over FG will be used for the prediction at the next state of model identification to further improve the estimation accuracy without requiring unacceptably high computation effort.