I would be very interested in ways to modify the AI acceleration. Do you have any idea?Klaabu wrote: The same old rocket-like acceleration if used as AI. Can't they really learn from others or at least ask some help from RSC? As much as I understand they are RSC-s official partners so it's should not be so hard. The best example of good physics (as good as possible in RW) is Class 380 EMU released by Totalize Media. I suggest to run a little test - place Class 380 and German BR-120 with 4 coaches side by side on parallel tracks as AI and watch how do they accelerate. If You add 8 coaches to the German train acceleration remains almost the same.
As far as the comparison is concerned, do both models have the same power? I have no idea about the 380, but it could well be that it has two motor units, each with half the power, to make it work in RW when you drive it. And the AI just looks at one motor unit and thus uses half the acceleration. Just one of many possibilities.
My understanding is that the "AI" does some approximation of the correct acceleration, but I am not sure based on what. And then, they miscalculate by some factor. In my experiments, train length was an influence factor, and the factor by which the acceleration was overdone was between 3 and 4, not exactly constant, but quite so. However, that was for KRS, I have not got around experimenting more with RW, and actually, bug fixing by experimentation is not on my highest priority.
