I seriously wonder if they haven't already done it. The increasing complexity of crop patterns might be down to competition between groups of students, but it does also coincide with this kind of technology. Older crop-patterns were mostly based on circles, which can done with pegs, ropes and planks. I hypothesise (though am too lazy to survey the evidence) that the more complex non-circle patterns have only arisen since these kind of robots became feasible.
One of the reasons I asked about GPS earlier is that if a lawnmower-type robot were programmed to go, say, eighteen inches forward and then turn left, its driving wheels/tracks might slip on the crop, and the distance would not be read accurately enough for such complex patterns. GPS would solve this if it could provide fine enough detail.
The other reason I suspected GPS was that in some example I've seen, the crop circles have been on sloping ground, and have actually been elliptical, only appearing circular from directly above.
