When I show up at JPL, NBC is already there, filming Jeff and Mark as they simulate an Opportunity traverse in the testbed. Or, more precisely, outside the testbed, on a pile of dirt trucked in that day to simulate a crater rim. Our faux Opportunity is doing a trenching experiment -- holding five wheels still and turning the sixth, which causes it to dig a little way into the dirt. Then it turns back and forth a little, which compacts the soil on the walls of its trench, and digs some more. This works pretty well on level ground, but the problem with doing it on a slope, they're discovering, is that the rover slides downhill as it does this, so instead of digging a deep trench in one place, it leaves a sort of broad but shallower trail in the downhill direction. Which isn't necessarily fatal to the whole plan, but it means that when they're on the crater rim, they probably can't dig as deeply as they normally would. It's unclear, though, how much of the slipping is due to Earth gravity and the fact that our soil is wet, which is not a problem on Mars. Not that NBC cares. They're just here to get their visuals, and this experiment is pretty cool to watch.
When they're done with this, Mark and I get to work simulating the alternate traverse we planned out yesterday, which roughly goes 4m forward, turns 45 degrees clockwise, then does another 5.5m. One unexpected problem is that we don't have enough room -- the inside of the testbed has an obstacle course set up for the other rover, and we don't want to disturb that. We find a path we can use, move some camouflage netting aside, and drive the rover inside -- only to discover that there's not as much room as we thought there was. So we have to drive back outside and improvise a path that starts on the slope and goes downhill, then turns and leads out onto the asphalt.
Like everything else you do in the testbed, this takes way more time than it should. I show up at 5PM, but we don't really get started until 8PM, at which point we're thinking we'll be done in maybe another hour.
Yeah, right. The testbed is not only a time warp, it's a reality distortion field worthy of Steve Jobs. You're always about an hour away from being done, and you never learn.
Several hours later, we've discovered a couple of bugs, and I go get us food while Mark fixes them. Several hours, and three or four traverse attempts, after that, I look at Mark and say: "It's two o'clock in the morning, it's cold outside, and I'm standing on a pile of dirt, baby-sitting a robot." "And having a great time," he adds. And having a great time.
Each attempt, of course, we get a little farther. But the rover only goes 5cm per second, and that's when it's moving. Between moves, it has to stop and think, and one of the bugs we discover makes the stop-and-think part take 3.5 minutes instead of 30-90 seconds. Which triggers a timeout, which makes us start over. And every time we start over, we have to drive the rover back to its starting position before we can try again.
Eventually, sometime after 4AM, we get it done. A successful drive! For the trip back, we turn the rover around and test the original White-Boat drive, a short backup followed by a loping 4m arc that turns 25 degrees over its total distance. This goes a little faster, since it relies on longer moves and we've solved the slow-stop-and-wait problem, but it still takes almost another hour to get inside. After that, we still have to park the rover and shut down the test bed.
It's after 6AM when I leave. I have no idea what's gone on with the real rovers while I've been here. I realize that I've just spent more than twelve hours standing outside on a cold night; the moon had just risen when I arrived, and it's long since set. I ought to be freezing and hungry and tired and sore. I'm none of the above.
 "Why" -- I can hear you asking -- "was there camouflage netting in there?" Because without the camouflage netting, the rovers' vision system sees the walls of the testbed as a hazard, and frequently prevents driving when we don't want it to. With the camouflage netting, the rovers' vision system doesn't see the walls at all. As a project, we accepted the risk that we might endanger the rover if we found camouflage netting on Mars.