Sensor Repositioning to Improve Undersea Sensor Field Coverage

The growing demand for persistent underwater surveillance has led to a need to increase reliance on undersea distributed sensor networks for undersea target detection, classi fication and tracking. While tremendous progress has been made in the technology of small, relatively inexpensive sensors over the last decade, progress has lagged in the areas of sensor allocation and sensor management. How best to deploy and reposition sensors and small, unmanned vehicles (movable sensors) are important research questions that must be addressed to realize the intended use of these technologies. Realistic tactical sensor deployment scenarios do not provide the opportunity for a precise placement of sensors. Most likely, initial deployment will be somewhat random (e.g., deployment of sensors from a moving vessel). Additionally, sensors might have to be repositioned due to random sensor failure, degradation, drift due to ocean current or other environmental effects. While it is possible, through the use of geometric probability, to estimate of the coverage of randomly distributed sensor fields, optimum field coverage can only be obtained through the use of deterministic sensor positioning procedures. However, the initial randomly distributed sensor field can be used as a starting point for the optimal sensor placement. The same can be said for networks in which sensors have drifted out of position, experienced failures, or have (through random movement or collision) aggregated into clumps. Sensor redeployment might also be necessary due to changes in mission objectives. For example, improved intelligence might necessitate the need to reconfigure the network in order to detect the target of interest. This paper addresses various issues relating to repositioning of sensors in order to improve the coverage of the distributed sensor network. In addition to more traditional assignment algorithms, which minimize the total (equivalently, average) cost for moving all sensors, we consider various cost-based assignment techniques that aim to minimize maximal displacement. We argue that for some scenarios, especially small to moderate networks of sensors with limited fuel supply, the minimization of the maximal displacement is preferable to the solution of the more traditional assignment algorithm. The latter often produces results with relatively large costs for at least some of its assignments. This leads to diminished effectiveness over time for the sensor field. Since fuel supply is limited for these unmanned vehicles, we consider assignment procedures that will not deplete the vehicles' resources during the maneuvering phase. Finally, we compare the performance of several algorithms used to minimize the maximal cost associated with repositioning a field of movable sensors.