It doesn't seem to me that aerial combat is vastly simpler than ground combat. If a robot pilot can accurately identify friend and foe, and maneuver effectively to target the latter, then likely we are not a lot farther from doing the same with an infantry robot.
I would claim it is vastly simpler. going from one position to another in a aerial vehicle is much simpler than navigating a cityscape, a swamp, or even simply an uneven field.
In the aerial domain, your environment is for all intents and purposes empty, and you operate over and above a surface which may have targets on it. As long as you ensure your manuevers don't intersect the ground manuevering is mostly point in the direction you want to go. aerial -> aerial is even simpler, as these engagements are long range and often significantly above the ground, so other than maintaining altitude you can treat the whole universe as empty except you and your target.
There are numerous videos of robots traversing rough ground now. None of them can approximate a human athlete doing the same, yet. But I remember when the same was true of the best chess programs versus humans, and now the one on my phone can beat an international grand master.
How do you identify civilians on the ground? Are they as easily identifiable at a glance like objects in the air? Do they all carry transponders to identify themselves wirelessly?
We’re certainly far away from effective solutions for either but one problem is much easier to solve.
