Design and implementation of a biomimetic robotic fish with 3D locomotion

This paper investigates the design and implementation of a biomimetic robotic fish with 3D locomotion. The multi-joint body structure of the robotic fish is designed to simulate the carangiform locomotion and the pitching is controlled by a “two-tanks”-based diving system. A GIM-based biomimetic learning approach is adopted to learn and to regenerate carangiform fish swim patterns to the robot. By virtue of the scaling properties, the learning approach is able to generate various 3D motion patterns by tuning parameters without arduous learning or training process. Finally, the performance of the robotic fish system is analyzed through experiments.