Mobile target localization through low complexity compressed sensing with iterative alternate coordinates projections

In this paper, we evaluate the potential of several compressed sensing (CS) techniques for localizing mobile targets within a wireless sensor network. First, we point out the limitations of popular algorithms enabling greedy s-sparse signal recovery, such as the recursive least-absolute shrinkage and selection operator (RLASSO) or the simultaneous orthogonal matching pursuit (SOMP). Then, we adapt the previous methods, making use of their non-binary outputs as soft information while accounting for the presence of a mobile target over a 2D grid. We also reformulate the localization problem by considering separable coordinate-wise CS dictionaries and accordingly, we introduce a new iterative gradient descent based solver relying on alternate coordinates projections (IACP). In comparison with conventional approaches, the latter CS solution benefits from arbitrarily fine spatial granularity at very low computational complexity. Finally, we show how successive restrictions of the search area under mobility can contribute to achieve even better localization performance and lower complexity for two of the proposed CS algorithms.