Lower-limb prostheses are used to restore amputee's walking. Monolimb is one of the designs referring to socket and the shank being molded into one piece of thermoplastic material. Appropriate shank flexibility of a monolimb can improve gait of an amputee. However, during the fabrication, the variations of design variables are inevitably produced which may lead the unexpected shank deflection and directly influence on gait efficiency of an amputee. This paper presents a robust design procedure for improvement of quality of the monolimb by simultaneously minimizing performance variations caused by variations in design variables and bringing the mean value of performance on target. The robust design procedure embodies the integration of response surface methodology with genetic algorithms. Response surface models are developed for the responses of monolimb as functions of design variables over the region of interest and genetic algorithms are employed to find the robust solution. A robust design of monolimb is performed for an amputee subject and the results show that the robust design can design a "robust" monolimb which provides specified performance targets that are minimally sensitive to the variations of design variables. This indicates that robust design may have the potential application in improving the quality of the prescribed prosthesis.