Does it actually include touch and go like driving with the driver as decision maker? I expected that every move would be scrutinized by a team of specialists?
In a nutshell, one rover planner (that's what drivers / arm operators are called) writes the sequence of commands, a second rover planner checks it and then walks through it line-by-line out loud with another ~2 rover planners watching. Then the second rover planner walks through the whole thing again with many other teams listening in and asking questions.
The first rover planner also consults with a surface properties scientist during development of the drive command sequence.
So, if things go wrong it's never really any single person's fault. But still, that first rover planner drafting the initial version of the command sequence bears a fair amount of responsibility.
Just wanted to say that this is the kind comment chain I come to HN for, when a Mars rover driver comes in with a relevant and amazing anecdote. Absolutely love this. Thank you foobarbecue!
The autonomous driving capabilities on Curiosity (known as "guarded" and "autonav" modes) are quite limited and haven't been used recently. They don't work in rough terrain, and we've been in rough terrain for a while now. This means that we can only drive to the end of the 3D meshes generated from images from the previous parking spot. When the terrain is acceptable for autonav we can usually rely on orbital imagery and drive farther.
We do set limits on suspension angles and tilt of the rover. So we'll say "stop driving if you are tilted more than 20 degrees," "stop driving if your differential angle goes above 7 degrees," etc.
Does it actually include touch and go like driving with the driver as decision maker? I expected that every move would be scrutinized by a team of specialists?