Distributed and centralized multisensor detection with misaligned sensors

Abstract

We consider the problem of multisensor detection in the presence of misalignment. We assume that the region that is covered by the sensors contains subregions that constitute blind spots in the sensors' fields of view. For analytical simplicity and numerical convenience, we consider the two-sensor case only, and describe the misalignment mathematically using a model that we have developed earlier. Preliminary assumptions involve a known geometry of the regions covered by each sensor and symmetric coverage. We formulate and analyze the distributed decision problem in the presence of misalignment when the sensors transmit only local decisions to the fusion. Different combining rules are considered at the fusion and compared with a centralized fusion scheme. Numerical results in the Gaussian channel indicate that for two sensors and under the imposed assumptions, only the OR combining rule at the fusion results in performance that degrades gracefully as the coverage factor decreases. The performance of the fusion under the OR rule is comparable—although inferior—to the performance of the centralized scheme. However, the AND combining rule yields very poor performance that degrades rapidly as the coverage factor varies.