DISTRIBUTED SAR DATA PROCESSING AIDED BY MACHINE LEARNING

The distribution of the space key resources, usually concentrated in a single, large, and complex satellite, can be shared among small-sized and simpler systems, thanks to the proper combination of the signals from each single node of the swarm. A proper sensor positioning, and on-ground data recombination allow the azimuth ambiguities to be cancelled out, guaranteeing gain in terms of signal to noise ratio. However, sensors synchronization, position control and trajectory knowledge errors are significant impairments which may make ineffective the recombination algorithms since the error contributions are hard to estimate. In this paper, a Deep Learning based approach is proposed for performing the recombination of data coming from a swarm of satellites and where no a priori knowledge error is exploited. The proposed system provides promising results in terms of ambiguity rejection.