Peter used to consult/collaborate with my lab. He was a proponent of moving remote sensing computations closer to the sensor ("edge computing" these days). You

Peter used to consult/collaborate with my lab. He was a proponent of moving remote sensing computations closer to the sensor ("edge computing" these days). You could definitely make the intellectual case for this approach. In cases where there's latency or costs associated with moving data to central computing, this makes sense. (In our case, it was space-based sensors so you could make the case that way.) But AFAIK this style of processing was never systematically adopted in any space-based processing system, although many such systems (like radars) have ad hoc data reductions performed in hardware close to the sensor. Thanks for providing that connection!