Nice article! I used to work in this field and there is an insane amount of cool tricks you can do if you have enough signal processing available. The first (obvious) one is that, since you can switch beam direction as quickly as you can switch radio frequency (ie thousands of times per second) it is no longer necessary to give each target equal illumination time. It is for example possible to mostly focus on the missiles inbound to your vessel at mach 3 while still not losing track of all the other traffic in the area. You can't really do that with older mechanical antennas since the inertia would tear the assembly apart.<p>The second is "colored space" radar, where you arrange the phases and wavelengths in such a way that the same transceiver array can generate multiple beams with different frequencies in different directions, giving true parallel beamforming (not merely concurrent as in the first example).<p>Finally, though every textbook introducing these concepts uses transceivers arranged in a flat plane, that is not actually required. For example, the F-35 uses "conformal antennas" that are shaped like the rest of the airplane and uses a phenomenal amount of signal processing to convert the resulting signals back to "as if" the antenna had been a flat plane.<p>I have always wanted to experiment with these things and rectennas to power small UAVs from afar so that they never have to land, but never got around to it.
The folks at DARPA want to investigate the use of Machine Learning math and tools to control and process antenna arrays. See the Tensors for Reprogrammable Intelligent Array Demonstrations (TRIAD) program, referenced here:<p><a href="https://intelligencecommunitynews.com/darpa-releases-triad-aie-notice/" rel="nofollow">https://intelligencecommunitynews.com/darpa-releases-triad-a...</a>