Gaussian Processes for Active Sensor Management

In this paper we study the active sensor management problem using continuous optimal experimental design (OED) framework. This task comprises the determination of allocation for a limited number of sensors over the spatial domain and the number of repetitive measurements in these locations in order to improve the overall system performance. We present a principled approach to active sensor management with repetitive measurements for Gaussian Processes (GPs) using a generalised D-optimality criteria and soft margin constrains. The resulting optimum of the convex optimization of the optimal experimental design for GP is generally sparse, in the sense that measurements should be taken at only a limited set of possible sensor locations. We demonstrate the use of our method on arti¯cial dataset.