These comments are from Ingo Cyliax, who drove part of the first DPG route on October 16/17, 2004. See also Ingo's photographs
My initial reaction of the first leg of the course are "we're crazy". The first leg littered with washes and ruts, and the hill it went over was pretty hard to navigate. I think we figure out where Team Red stopped, and there were worse sections after that.
We couldn't figure out the section of the course between I40 and I15, since there were a lot of little roads and I didn't have a GPS sync'ed moving map to help. Also, there were some private properties we couldn't go through.
We did a section from kingston road to the end of the course, which was rugged, but not as bad as the first section. Lot's of rocks, and some deep washes. The road through them was filled up, including the wash at WP2120. I can see how after a good flood the road will wash out and away and you end up with something that looks like the course describes in the waypoint list, but in our case it went straight through.
On the way back we went from SR127 through to I15 (backwards), this includes a few mountains with tight switchbacks. The road is pretty good in most places, however, in one section, there were a lot of sharp rocks and we lost a tire. There were also a couple of sand traps on the road in other sections.
One note, you can't go fast in many of the tight places and definately not on the first leg of the course. In general, it was hard to go any faster than 20-30Mph, except some places where the road is packed. However, even in those sections there were ruts and washouts that occur without warning, i.e. they blend into the road visually.
while I was driving the course I tried to keep in mind what I was reasoning, in order of concern.
Also, here are some thoughts.
This image [is] one of the best. See if a vision algorithm can make sense of this. Hint: there are three roads here. BTW, with a camera that has high dynamic range, you should be able to make out some details in the shadow. When it's sunny, there is enough scatter to illuminate the shadows. On overcast days, there are no shadows and the light level is not much lower than in the shadows on a sunny day. [you could probably simulate the dynamic range of a video camera, by assuming you get about 6-8bits of levels. the camera I was using is supposed to be getting 10-12bits/color, but probably less than that.] You can make a historgram of the camera image and adjust the level to get good distribution in the areas of the image you are interested in. Another way around this, is to have two (or more) cameras with exposures (levels) adjusted to give a continous range. I.e. one camera is for looking in shadows and the other for looking at non-shadows. Fill lighting for shadows will be a problem. Laser might be OK, if they are bright and you use a filter on the camera to only pass the laser color. E.g. infrared.
This is a picture of a wash. what's interesting is, that it was a pretty deep wash and the picture doesn't really reflect how bad it was. You can get kind of an idea of it by looking at the features off the side of the road.