Improving sensor-fusion with environmental models

This paper presents an algorithm to improve sensor fusion results in outdoor WSNs using environmental models to redefine periodically which sensors to use and the amount of weight for each particular sensor in the fusion solution. Using daily forecast simulations of the monitored environment dynamics, clusters of sensor nodes sharing data correlation can be defined, for priority sensor selection in a fusion algorithm. The strategy was validated in an operational aquatic sensor network comprised of several sensors scattered over a multiple square kilometers area. It allowed the use of geographically separated nodes with confidence and sometimes even in alternative of closer nodes, unlike most fusion approaches that use quasi-redundant information provided by nearby sensor nodes. This approach represents a major step in the creation of a framework to assess the validity of the monitoring data in WSNs.