To solve the problem that there is no basis for choosing control signals and sensor information for central pattern generator (CPG) control of snake robots, a cyclic inhibitory CPG control method with mechanical oscillators is proposed. Firstly, the mechanical oscillators reformed from snake robot dynamical equations are introduced into the cyclic inhibitory CPG model. Secondly, an improved Matsuoka neuron is proposed so that the neuron and mechanical oscillator can be expressed in a uniform way. Thirdly, the relationship of parameters in the cyclic inhibitory CPG model with mechanical oscillators is illustrated, and the relationship expression of the control signal and sensor information for snake robots and the CPG states is given. Finally, the proposed method is verified with simulation, and the simulation results are analyzed. In this method, there are clear definitions for the control signals and sensor information of snake robots, and the computation complexity of CPG is decreased because neuron computation in CPG is replaced with physical structure of mechanical oscillators.