Abstract: A central pattern generator (CPG) based locomotion method for a snake robot by Hopf oscillators is developed according to different gait characteristics of biological snake. Firstly, coupled Hopf oscillators with the feature of nonlinear limit cycle are selected to construct a chain network model that can achieve serpentine locomotion and sidewinding locomotion. Then, a simulated snake robot is created by a dynamics simulation software and is driven by the outputs of the oscillators in the model which are able to generate control signals for distributed redundancy joints of the snake robot. A successful implementation of the two kinds of gaits in simulation environment is demonstrated and the relation between CPG parameters and the forward velocity of the robot is also discussed. Finally, experimental results on a real snake robot confirm the effectiveness of the proposed method to achieve different gaits of the snake robot.