SPRTs in sensor networks with mobile agents

We assume a detection network engaged in the task of sequentially testing simple Gaussian hypotheses. The network architecture follows the SENMA paradigm but, as opposed to the usual assumptions, the mobile agent (rover) implements continuous-time polling. This, along with a Poisson random field model for the sensor locations (which are completely unknown to the rover), allows us to avoid channel collision concerns. The remote sensors employ Wald sequential tests (SPRTs) so that, when any is eventually polled, it may have reached a (local) decision. The rover also implements an SPRT using as observables the local decisions (amounting to some counting processes). The idea behind is that of reducing the time needed to end up with a final decision by the rover, with respect to similar, but nonsequential, schemes. The local SPRTs, moreover, ensure energy savings in terms of communication needs.