The number of sensing devices needed would be very high to identify the signals or targets at a fine resolution. The general sparse distribution of signal sources makes it possible to apply the emerging sparse sensing (CS) technique to recover the signals. Although promising, the CS-based schemes cannot work efficiently in a practical environment due to several reasons: 1) The number of targets may be unknown in advance and can vary over time; 2) The signals often distribute unevenly in the monitoring space; and 3) The computation complexity for signal recovery is very high when the signal vector is large to achieve high resolution monitoring in a large domain.

We propose to adapt the sampling distribution as well as detection resolution based on the knowledge on past measurements to more significantly improve the sensing performance. Specifically, we propose to develop two tiers of learning-based feedback control schemes, one performed at the local sensor level and the other performed at the larger scope network level for more efficient collaborative sensing among sensors. Specifically, we propose to 1) develop a virtual zoning scheme to significantly improve the detection performance with given sampling data; and 2) design two sequential sensing schemes to gradually increase the number of samples to improve the sensing accuracy. The two schemes will control the physical sensors to adapt their behaviors or detection resolution based on the results they learn from the previous rounds of processing or sensing to reduce the uncertainty and ambiguity for more accurate detection in a dynamic and varying physical environment. The exploration of learning, rather than random sampling, would enable either of our algorithms to achieve higher sensing performance or significantly reduce the sensing resources compared to that achieved based on the conventional CS theory. Performance results demonstrate that our scheme can achieve significant gains in sensing accuracy, processing timeliness, cost reduction, and energy conservation.