A hierarchical neurobiologically inspired control framework for stable and energy efficient locomotion of bionic robotic fish

Balancing behavioral adaptability and actuation stability is crucial for biomimetic autonomous underwater vehicles (AUVs), inspired by fish sensory-motor systems. While biologically inspired computing methods have advanced rhythmic motion generation, it remains challenging to maintain smooth joint coordination and propulsion efficiency under varying aquatic conditions. In this paper, we propose a hierarchical control framework integrating a macroscopically partitioned Shallow Brain Model (SBM) with a multiple oscillator Central Pattern Generator (CPG). The superior SBM handles multimodal sensory fusion and interpretable decision making across six distinct functional brain regions, while the subordinate CPG ensures synchronized rhythmic actuation under hydrodynamic disturbances. We benchmark against a Leaky Integrate-and-Fire Spiking Neural Network (SNN) with optimized synaptic filtering. Extensive simulations and frequency domain analyses reveal that while synaptic filtering alleviates actuator jitter in SNNs, it inherently introduces phase latency and peak amplitude attenuation. In contrast, the integrated framework directly yields continuous and high fidelity commands, achieving an exceptional Signal to Noise Ratio (SNR) of 40.4 dB and reducing the Cost of Transport (COT) by 27.9% in simulation and 13.96% in physical trials. The results support a biologically inspired, energy-efficient control architecture for robotic fish under the controlled and naturally disturbed aquatic conditions examined in this study.