A cooperative deployment strategy for optimal sampling in spatiotemporal estimation

This paper considers a network composed of robotic agents and static nodes performing spatial estimation of a dynamic physical process. The physical process is modeled as a spatiotemporal random field with finite spatial correlation range. We propose a distributed coordination algorithm to optimize data acquisition across time. The robotic agents take measurements of the process and relay them to the static nodes. The static nodes collectively compute directions of maximum descent of the estimation uncertainty, and relay them back to the robotic agents. The technical approach combines tools from geostatistics, parallel computing, and systems and control. We illustrate the soundness of the algorithm in simulation.