A bionic robotic fish is a fish-shaped robot that exhibits exceptional movement efficiency, rapid swimming speed, and impressive maneuverability. Nevertheless, the motion control of robotic fish is exclusively based on oscillatory motion in water, which lacks biological reasoning and interpretability in motion control, thus unable to meet the demands for high motion efficiency and adaptability. This work proposes a hierarchical motion control method aimed at enhancing the motion efficiency of robotic fish. The method utilizes spiking neural network (SNN) and central pattern generator (CPG) to optimize the control system, and introduces Spike-Timing-Dependent Plasticity (STDP) learning rules to adjust synaptic weights, thereby enhancing the model’s adjustability and learning ability to better mimic synaptic plasticity between biological neurons. The established hierarchical control strategy is compared and verified in terms of speed, stability, and action fluency through simulations and experiments of different control methods under different spiking neuron topologies. To summarize, this research offers a scientifically feasible alternative for the motion control of robotic fish.