LSTM-Based Learning Control for Biomimetic Robotic Fish Swimming

To precisely manage the location and trajectory of a biomimetic robotic fish moving in water, it is crucial to comprehend its pressure and derive the related dynamic equation. In the process of dynamic modeling, the movement of robotic fish is affected by its shape, mechanical structure, and complex fluid force, making it difficult to precisely regulate the robotic fish's movement. Consequently, this article employs an LSTM (Long Short-Term Memory) neural network to discover the variations in robotic fish joint angles. Meanwhile, simulations and experiments of LSTM-controlled bionic robot fish swimming straight and turning are presented. These simulated and experimental results validate the method's efficiency.