Optimizing Potential Information Gain from Satellites over Space and Time for Heterogeneous Information Fusion

Given a certain number of areas of interest on Earth, the objective is to maximize the information gained from these areas using different sensors on available satellites. Various constraints exist, such as weather constraints (cloud coverage might prevent images from being taken), energy constraints, and feasibility constraints. We use a spatial interpolation method called Kriging for data fusion to collect information from cells other than those from which data is directly collected. Using Kriging, we can calculate the estimated prediction error in each cell. A cell is “confidently covered” if this estimated error is below a certain acceptable level. Since we have limited energy, we do not always need to transmit high-resolution data back to Earth. We create an optimization model that decides what data should be collected, from where, and the level of fidelity so that the number of “confidently covered” cells in high-interest areas is maximized. Since Kriging is not a linear process, we also develop a linear approximation for Kriging, use it in a secondary Mixed Integer Linear Program (MILP) model, and compare results.