Do the rovers use visuals to navigate at all aside from sending images to Earth, analyzed and then commands returned?
I think nowadays the drivers of Curiosity rover use images received from Mars to make plans for the rover's destinations and her path to the destinations. After the plans are made, the drivers send specific and detailed commands to the rover. The rover thus drives autonomously.
There is onboard software called "AutoNav" that allows the rover to drive some distance without detailed instructions. I think typically the first part of a drive uses detailed instructions from the ground (oddly called the "blind drive" part), and then the rover may drive some additional distance using AutoNav. The idea is the last part of the drive may be too far to have been seen well enough for detailed instructions.
From what I have learned, so many systems have failed on the rover that the operators have had to learn new ways of controlling it.
Cant remember where I found that.