Underslung girder bridges are effective structural systems formed by a rigid beam supported from beneath by an arrangement of struts and stays. The optimal bridge scheme can be sought tuning a large set of parameters of differing character. In the development of bridge design, however, structural engineers usually use uniform arrangement of struts and stays. This paper suggests employing in the design process the structural optimization and shows optimal schemes of underslung girder bridges for different spans and deck types. Mathematically, the optimization of the bridge structure is a mixed topology-shape-sizing optimization problem with a variable number of design variables (some variables are integer). An evolutionary algorithm with variable-length genotype and special operators for the mutation, crossover, and genome repair was used as the optimization engine. Moreover, the stochastic findings of the algorithm thereafter can be improved using consecutive pattern search.