This paper, inspired by neuroanatomy, proposes a parallel computing architecture named the Shallow Brain Model for the swimming control of bionic robotic fish. The model simulates the functions of the hippocampus and cerebellum in organisms, and realizes the movement and obstacle avoidance functions of the robot fish by adopting the method of regionalized brain processing.The shallow brain model uses a recursive fractal neural network structure to simulate the computational process of the biological nervous system. It can process data from different sensors and generate motion control signals in real time based on the data. The experimental results validate that the model is capable of accurately simulating the swimming behavior of the robotic fish and possesses strong generalization ability. Moreover, this study systematically analyzes the performance of the neural network, validates the model’s adaptability in dynamic environments, and presents a novel approach for the swimming control of bionic robotic fish, demonstrating significant potential.