Efficient Spatial Signals Monitoring From UAV Observations Using Dual Stochastic Gradient

Dual stochastic gradient algorithm is proposed and discussed for efficient detection of spatially correlated signals using unmanned aerial vehicles (UAV). The spatial signal is reduced to a number of its contour lines for efficient modeling purpose. A computationally powerful data fusion center (DFC) estimates a batch of suitable contour levels and dispatches the UAV(s) to track each contour line and report their tracking coordinates to the DFC for spatial modeling. Each new batch of estimated contour levels is based on the most recent spatial signal estimation. The new additional number of contour levels is estimated based on a stochastic gradient algorithm. To reduce the flying distance of the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\operatorname{UAV}(s)$</tex>, those new contour levels that are too close to any of the previously paced contour levels are dropped at the DFC in a second stochastic gradient process. The performance evaluation of the algorithm based on computer simulations presents the significant efficiency of the proposed dual stochastic gradient algorithm. The spatial modeling error is used as performance measure and the total flying distance is introduced as the algorithm's cost.