Joint sensors-sources association and tracking

This paper considers the problem of tracking multiple sources using observations acquired at spatially scattered sensors. Kalman filtering and smoothing techniques are combined with a sparse matrix estimation framework. A pertinent normone regularized minimization formulation is proposed that jointly searches for source-informative sensors, associates sources with sensors and tracks the unknown sources. Block coordinate descent techniques are used to recover the unknown sparse observation matrix, and subsequently obtain source state estimates. Numerical tests are provided to demonstrate the potential of the novel approach to identify the source-informative sensors and accurately track the field sources.