When NASA first proposed a single-SRB-based launcher, one of the issues that jumped out immediately to many familiar with vehicle design and Shuttle design was roll control. As designed for the Shuttle, there are two SRBs, both of which can gimbal the engines. This allows roll control of the Shuttle stack by gimbaling them in opposite directions. But when there's only one, the engine gimbal provides pitch and yaw control, but there's no way for it to control roll.

There are two potential solutions to this--to modify the SRB itself to add roll-control thrusters, or to incorporate them into the new upper stage. The latter has the disadvantage of oversizing the roll-control system for the period after stage separation, which adds weight and affects performance, but it simplifies design by requiring only one system.

In any event, the concept seems to be in trouble. Now this certainly isn't a show stopper, and issues like this are inevitable in the development of a new launch vehicle, but it's just one more demonstration of the fact that deriving a new launcher from existing pieces isn't as easy as has been advertised by many, both within and out of the agency.

[Late morning update]

Gary Hudson emails one other option:

There is a third possibility: let it roll. Depending on the rate and duration, it may not be a problem. Some current vehicles do this (Taurus, for one) and we are planning a subset of it for the AirLaunch QuickReach. In our case, we have a Stage Two roll thruster but its purpose is to limit the rate, not hold a specific roll attitude. Makes for a much small thruster. It is later used as part of the normally smaller sized Stage Two attitude control subsystem.

Errrr.... where in the article linked does it say something like "the concept seems to be in trouble?"
" the agency is looking at potential operational risks associated with each of the two engine approaches for the Ares Upper Stage roll control engine." This sounds more like basic engineering than "trouble."

There is a third possibility: let it roll. Depending on the rate and duration, it may not be a problem.

Yes, but humans have a much lower tolerance for spinning around than do machines. Just ask Mike Melvile :P

"Yes, but humans have a much lower tolerance for spinning around than do machines."

The rates for this "let it roll" option are typically fairly small, say 10-25 deg/sec. Lot less than the normal Shuttle 180 deg roll done after clearing the tower.

The problem with letting it roll, from what I have been told, is blood flow effects on human crew. You want to be head down in high g situations.

"The problem with letting it roll, from what I have been told, is blood flow effects on human crew. You want to be head down in high g situations."

You are head down, either way you look at it. Orion is designed so that the crew is sitting with their heads lower than their hearts under axial thrust. And if you are rolling, the (minor) G component of that is also forcing blood to the brain. Finally, in much higher G re-entries we also roll to bank, and thus obtain crossrange.

Has anyone heard if they considered rollerons?

http://science.howstuffworks.com/sidewinder4.htm

They work well for Sidewinder missiles.