Self-organizing cooperative sensor network for remote surveillance

The capabilities of unattended ground sensors (UBSs) have steadily improved and have been shown to be of value in various military missions. Today's UGS are multifunctional, integrated sensor platforms that can detect and locate a wide variety of ground-based and airborne targets. Due primarily to cost and size constraints of these UGS, they have not been widely used for law enforcement surveillance applications. As an alternative to a single, monolithic sensor platform, remote surveillance may be possible with smaller, less obtrusive sensors that work cooperatively together as a network. The objective of this study was to develop algorithms that can optimally organize and adaptively control a network of UGSs in order to achieve a surveillance mission. In the present study, the sensor network, a random distribution of sensors over a surveillance area (emulates airborne sensor deployment), determines an optimal combination of its sensors that will detect multiple targets and consume the lease amount of power. This problem is considered a multiobjective optimization problem to which there is no unique solution. Furthermore, for a linearly increasing number of sensors, the combinatorial search space increases exponentially. To reduce the search space, a novel clustering method was developed based on whether the sensor can sense the target rather than on similarities between the sensors. A genetic algorithm (GA) was used to obtain a quasioptimal solution for the sensor combination problem. To evaluate the effectiveness of the optimization, figures of merit were developed that are applicable to a sensor network tasked with a surveillance problem. Software-simulated data was used to test software implementation of the clustering, optimization and figure of merit functions. The clustering method reduced the search space by an average of ten orders of magnitude. For a sensor population of 100 sensors that was tasked to detect 24 targets, the GA was able to select optimal sets of sensors for detection and minimization of power consumption. The results demonstrate the feasibility of optimally configuring and controlling a network of sensors for remote surveillance applications.