Study For The Multi-joint Robotic Fish With Autonomous Obstacle Avoidance Based On The GIM Biomimetic Learning

Human has increased the exploitation of marine resources with the rapid growth of the population and the increasing scarcity of land resources. As a new-type underwater robot, combined with fish propulsion model and robot technology, biomimetic robotic fish has advantages in high efficiency of propulsion, maneuverability, strong concealment and low noise. So, it's full of economic value and military significance to develop new-type biomimetic robotic fish. In this paper, the mechanical structure, movement patterns generation theory and obstacle avoidance algorithm were investigated deepgoingly and systematically. The main research contents are listed as follow:1. 3D printing technology was applied to the design and manufacture of robot fish. Biomimetic robotic fish must have good water resistance and balance as a kind of robot that works under water. Due to the problems of materials, assemblage, sealing and other aspects in the traditional manufacturing process, robotic fish has problems of imbalance and water leakage in the course of swimming. 3D printing technology can print any complex shape of the product. Using 3D printing technology to print the machine fish shell, connection parts and the sealing parts, can achieve seamless connection between components, so that the robotic fish has compact structure, stable performance and better waterproof performance.2. Using a biomimetic learning method based on General Internal Model(GIM) to capture and study the movement of Carangiform fishes. Firstly, the real fish behaviors recorded by video system are used to construct the training samples. Three basic swimming patterns, “cruise”, “cruise in turning” and “C sharp turn”, are extracted from these captured data. Then, GIM is employed to learn the swimming patterns of Carangiform fishes. Based on the approximation ability and the temporal-spatial scalabilities of GIM, robotic fish can learn, modify and regenerate the similar swimming patterns of the real fish. Finally, the experiment was carried out using 3D printing machine fish. The experiment results verify the effectiveness of the proposed biomimetic learning approach.3. A kind of autonomous obstacle avoidance algorithm was established on the basis of the adaptive neuro-fuzzy control, and biomimetic robotic fish can effectively avoid obstacles in the course of swimming. Robotic fish obtains environmental information including the obstacles distance and direction information by infrared sensors. Due to the need of multiple sensor inputs, and the sensor data acquisition accompanied with error and uncertainty, the traditional obstacle avoidance algorithm can not be applied to robotic fish. Fuzzy control does not need to establish a precise mathematical model, which has strong fault tolerance and can effectively control with incomplete information. In addition, through the study of samples, we obtained membership functions and inference rules of fuzzy system, avoiding redundancy and errors arising from manual rules.