One assumption I made on early chain designs was that the sphere would sit directly atop them. In another configuration, the Delta (or triangular) version the chains are now sitting at an angle to the sphere. Chains are extremely strong but not designed to handle twisting very well, so this arrangement could cause potential failure or lower efficiency.
One solution was to use a ‘double chain’, which are fairly common. This means using a double sprocket, which are heavier and more expensive. If we end up manufacturing a chain for the project then there are several possibilities. I created a rough model to show one such possibility: 
This would widen the base, but also requires longer components. Perhaps not ideal.
A simple variant of this is to use two rollers instead of one. This increased surface contact doesn’t increase traction but spreads the wear on the rollers… with an additional cost for more washers and bolts.
These two solutions perhaps reduced the problem of chain twisting but probably at a quite high increase in cost. If we’re going to manufacture a chain then a single width version would always be best. Now perhaps you’re screaming at the screen the obvious solution, but this is an issue that we only just realised was a potential problem 🙂
Single width chain, double rollers. It has a slight increase in cost but we can go back to using standard sprockets.
The chain is by no means ‘locked’ into the design, and we’re taking another close look at omniwheels, but it’s good to have more options.
