Opportunistic Sampling for Random Field Estimation with M2M Capillary Networks

Abstract:
In this paper, we address the problem of the estimation of spatial random fields from the samples collected by a number of sensing devices in a capillary M2M network. In order to preserve batteries, sensors are often required to operate in a low duty-cycle regime. Hence, we ask ourselves whether opportunistically activating sensors, that is, only allowing sensors experiencing favorable conditions in the sensor-to-gateway channels participate in the estimation process, might be beneficial in terms of the resulting distortion. To answer this question, we derive closed-form expressions of the optimal encoding rate and the optimal number of sensors that, under a total bandwidth constrain and for a prescribed duty cycle, minimize the distortion in the reconstructed random field. The proposed opportunistic sampling scheme is then analyzed by means of computer simulations and its performance assessed and compared with that of other systems with random (i.e. non-opportunistic) sensor activation. The impact of the pre-defined duty cycle on delay-tolerant and delay-constrained communication strategies is analyzed, as well.