The problem of designing a robust three-axis missile autopilot that operates under aerodynamic roll angle uncertainty is addressed in this article. A finite number of local state-space models over an aerodynamic roll angle envelope are developed as a multi-model to represent uncertainty bounds. Two design methods with multi-objective output-feedback control are proposed. In the first approach, a classical three-loop autopilot structure is slightly modified for the multivariable autopilot design. The optimal gains in the autopilot structure are automatically obtained by using a co-evolutionary optimization method that addresses competing specifications and constraints. In the second approach, the mixed H2/ H∞ performance criteria are guaranteed by multi-objective control synthesis via optimization techniques. Both design approaches are used in non-linear simulations with variations in the aerodynamic roll angle to provide satisfactory performance as a three-axis missile autopilot.