Direction finding

What possibilities exist for a node to coarsely determine bearing to another node in the mesh?
Do the possibilities change if one or more nodes are mobile?

The obvious option is for nodes to receive NMEA messages from collocated GPS receivers and share coordinates.  That is largely solved.  I am looking for options where GPS is unavailable or unreliable.

Hypothetical 1:
Several AREDN units have multiple (though not necessarily differential) antennas.  https://forum.gl-inet.com/t/rssi-direction-finding/18809
If we can determine which has a stronger signal, that is enough for coarse direction finding.  How do we get that differential received signal strength information to a program?

​Hypothetical 2:
Assuming a node has a single omnidirectional antenna, is there a way to programmatically create a null toward a specific bearing?
I'm imagining 3 masts equidistant from each other and a central omni antenna.  Each mast can either be laid down on the ground or erected.  When erected, it creates a null in the omni toward that mast.  Right?  Haven't experimented yet, assuming.  Erect each mast, take a signal strength reading, lay the mast down.  Repeat for each mast.  The mast direction with largest attenuation is the direction to source.  Is there an electronic way to produce the same effect with no moving parts?

Hypothetical 1.5:
Using the GL-AR300M16-EXT from Hypothetical 1, use onboard gpio to temporarily switch in some attenuation that reduces gain on one antenna at a time.  Whichever direction has the largest signal drop with attenuation switched on is the direction of source.

Of course, since the differential rssi isnt taken simultaneously on both antennae, there could be other environmental factors that temporarily influence rss.  Therefore we cannot judge from a single set of readings, but will need to analyze trends and correlate with known movements of the nodes.

This is not my area of expertise at all.  I have no experience,  If this is a waste of my time, please tell me, but also tell me why it wont work.  If you have other ideas for node direction finding with any AREDN hardware, I'd love to hear it.

Rotating a dish antenna is a well proven method.  It is certainly on the table.  But for my use on a cheap mobile robot, a solution with no moving parts is better and something close to omnidirectional pattern is preferred.
Does any AREDN hardware have a cardiod antenna pattern?  It would be nearly as good as omni for finding nodes in unknown directions, but still offer a null which is steerable by steering the whole robot.