This paper proposes a decentralized and coordinate-free algorithm to monitor spatial events in a dynamic scalar field. The events that are the focus of this paper are the appearance, disappearance, and movement of "peaks" (local maxima). However, ongoing work is extending the approach to monitor the full set of critical points (peaks, pits, and passes). Our approach is based on the gradient flow between immediate neighbors in a geosensor network. Experimental investigations demonstrate that our algorithm is scalable, with O(n) overall communication complexity (where n is the number of nodes in the geosensor network), and even load distribution. Further, our algorithm can improve the accuracy of the identification of peaks in a limited spatial granularity sensor network when compared with an alternative approach that does not account for granularity issues. The results of this research have wide application to decentralized monitoring of dynamic fields, such as changing temperature, pollution levels, or gas concentration.
↧
