Today started off in an interesting way. One of the remote scientists at NASA Ames was being interviewed by a French television crew, and they wanted to demonstrate control of Nomad. So NASA Ames took control away from the Carnegie Science Center for a couple hours, and went off exploring the desert. Much to their surprise, they happened upon what looked like an impact crater! There was much excitement and investigation, especially when a radial ejecta pattern was found, but when the field team arrived to collect some ground truth they discovered that this "impact crater" was one of a series of blast holes located just next to a road, with a new crater every 100 meters. Ah well.
We have been having problems with our differential GPS corrections. Differential GPS (d-GPS) works by combining two GPS units to get an estimate with higher precision. Our GPS units themselves have been working well, but the terminal server which connects the unit at the hilltop with the one on the robot has been acting up, periodically needing to be reset. Today we replaced that terminal server with some new equipment (a router), which we hope will eliminate the need to keep a person stationed at the hilltop. Preliminary tests indicate that this solution is working well; the router has not needed to be reset. In defense of the terminal server, its documentation does recommend that it be kept ventillated with clean, dust-free air, in medium humidity, with suitable AC power (not a small generator whose voltage drops 20V during radio transmission).
We have implemented some changes in the navigation system, which has sometimes been rather sluggish to respond to changes in steering direction. This system was last used by Carnegie Mellon researchers on an earlier robot called Ratler, built by Sandia National Laboratories. Nomad's steering system actually turns the wheels in the direction of travel (like the front wheels on cars), but Ratler drove using "skid steering". The advantage of skid steering is that it allows the robot to "turn on a dime," and change directions quickly just by slowing down the motors. The disadvantage is that it makes less efficient use of its motors, and results in unreliable odometry logs because the wheels often slide (or skid) along the surface. Since the navigation system was tuned on that previous robot, it tended to expect that Nomad would be able to turn very quickly, even at high speed, but unfortunately that is not the case. This is due mainly to the new steering mode, but possibly also to the very loose soil. So tonight we implemented some software changes that should improve its performance, by slowing it down until it achieves the newly commanded steering direction.
Today Nomad drove autonomously for at least 1646 meters. I say "at least" because during some of this time we had lost GPS data, so the only distance estimate I can make is to assume Nomad drove in a straight line between its last known point, and the point it was at when GPS was restored. Later on we will try to piece together a more precise estimate from the combined GPS, odometry, and paper logs.