I think if you were to use some sort of data acquisition-or your method for your testing, but use a treadmill lets say for a control device-you could really come up with some interesting data I think.
For example-if you set the teadmilt to run at 15 MPh (do they go that high?) and throttled the car to keep the car at a steady pace on the treadmill-you would simulate a more realistic load.
There would of course still be a certain percentage of error, but it would be more than close enough for our needs.
Ray