Optimizing the configuration patterns for heterogeneous distributed sensor fields
Thomas A. Wettergren, Russell Costa · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2012
When unmanned distributed sensor fields are developed for rapid deployment in hostile areas, the deployment may consist of multiple sensor types. This occurs because of the variations in expected threats and uncertainties about the details of the local environmental conditions. As more detailed information is available at deployment, the quantity and types of sensors are given and fixed, yet the specific pattern for the configuration of their deployment is still variable. We develop a new optimization approach for planning these configurations for this resource constrained sensor application. Our approach takes into account the variety of sensors available and their respective expected performance in the environment, as well as the target uncertainty. Due to the large dimensionality of the design space for this unmanned sensor planning problem, heuristic-based optimizations will provide very sub-optimal solutions and gradient-based methods lack a good quality initialization. Instead, we utilize a robust optimization procedure that combines genetic algorithms with nonlinear programming techniques to create numerical solutions for determining the optimal spatial distribution of sensing effort for each type of sensor. We illustrate the effectiveness of the approach on numerical examples, and also illustrate the qualitative difference in the optimal patterns as a function of the relative numbers of available sensors of each type. We conclude by using the optimization results to discuss the benefits of interspersing the different sensor types, as opposed to creating area sub-segmentations for each type.