How-to drive a moon rover?

As i follow the Part-Time Scientists over at their Facebook page i came to notice their "Remote Rover Experiment", this one: http://kck.st/LQQMwu and http://roverexperiment.com

What do you think is it realistic to test drive a lunar rover here on earth?
Also, what is to learn by this?

It looks like the main goal is to get as many participants to wear off their hardware. Maybe a simulation would be easier in this case.

Autonomy and durability, regolith resistance and radiation rating are some things I know would be important. Dust is a devil..

My take on a rover
A walking bot with a soft touch would work good, instruments could be enclosed inside a octopus beak style encloser with beetle wing style fold out solar would be cool, but a thermionic radation cell would last longer and not be as easily obstructed.
Think boston dynamics big dog coated in a epoxy shell with 8 legs, so if 2 break...... it has 6 etc.

A simulation only tests what you program into it, and only allows the problems and errors you allow. Real testing, even in a analog environment, is not so convenient and you learn more from the surprises.

Sit in it and drive it like a golf cart.....?

Except you have to do it while looking through a tube with a two second delay from what is really going on and with a four second delay to any steering or brake command.

If i had access to a reasonable amount of computing power and some good games designers i would cheat with 2 well spaced cameras i would use that live data(added to our existing sat data of the moon) to do a 3d sim of were i would be in 4 seconds and drive in the simulation with the real data doubled as part of a split screen and also send that data back to the rover so it could auto stop if the sim diverged significantly from reality ie it spotted a big hole in front of it not in the sim.

Very easy to do in a computer simulation where you create the hole. Not so easy in the real where first you have to image the hole, process it to gauge enough points to create topography, construct a map of it, and then decide if its large enough to present a hazard or not. Over and over again many times a second.
Also unless you are literally methodically crawling (like the Mars rovers have to do), you don't have the time to wait for a stop command to come. The remote itself has to be able to do it and then wait for what to do next. Otherwise it will be in the ditch before the ground-side computer or operator even knows about it.

Ironically, this artifact of teleoperation is what I think is going to drive manned space, instead of replace it. Autonomous and long-ranged remote operations are going to prove so unreliable, tedious, and unproductive when it comes to mining, manufacturing, and other large-scale activities, that you will need human operators within a light second or two of the 'bots to control them or to get them out of trouble. So while the humans will never be swinging a shovel down in the mines, they will need to be on the Moon, or out in the Belt overseeing the 'bots that do.

Time delay is actually very easy to simulate. Ever play a really laggy multi-player game?

On the other hand the Google autonomous car has now racked up 300k miles of driving without an accident (IIRC). That's better stats than a human. Admittedly, not so many rocks to avoid, but plenty of other obstacles.

Remember Murphy's Law applies particularly to autonomous devices.

Back to rover simulations, here you go:

MapleSim breaks new ground in HIL real-time simulation for planetary rovers

It's pretty much a press release/plug for MapleSoft, but interesting nonetheless.