[Gdeaver]

Yes, rough top and carpet are 3 dimensional and there is a grain to carpet too.
Leads to a complex model. Some deterministic testing would be the way to go. We tested both and went with rough top for our swerve. There are other options than conveyor belting. In 2010 we had alignment problems with our first swerve bot. We had extreme wear with the conveyor belting. To deal with this we tried bike tread. There is a high end bike store near us with over a hundred different tires in the store. Found 2 exceptional tread patterns that have monster grip in the forward reverse direction and slip in the side to side. We had allot of tread carcuses laying around so we ground the tread off down to the fabric. Used thick CA to bond the bike tire to the conveyor belt. Used some carpet thread to reinforce the edges and to prevent de-lamination. Fantastic treads. wear like iron and good traction when needed and slip when turning. Never tried it but, probably would be fantastic on 6 wheel base. Note that out of all the tires in the store only 2 patterns were good. It's allot of work but can go way beyond the performance of conveyor belting.