In the mammalian spinal cord, a central pattern generator (CPG) is supposed to be as neural networks. The CPG is responsible for producing locomotor rhythms periodically, and the rhythms are propagated down to flexor and extensor legs. So far, we have constructed a hardware CPG model from the mathematical description proposed by Rybak et al., and, using it, reproduced oscillation behaviors corresponding to some quadrupedal gaits of four-legged animals. However, when reproducing their gaits, we need to change the structure of the network among the hardware CPG models. In this technical report, we reproduced the oscillation behaviors corresponding to the typical gaits, "walking", "trotting", and "bounding" with the hardware CPG model proposed, and switched them using the only one parameter, voltage V_<MLR>. As a result, we constructed two hardware networks that can switch between the walking and trotting, and between the trotting and bounding.