Route Estimation Toward Autonomous Swimming of Robotic Fish

Underwater robots that imitate aquatic life, such as fish-like robots, have attracted attention for oceanographic studies from the viewpoint of morphological affinity for marine life. The existing studies concerning fish-like robots have primarily focused on the swimming mechanism and locomotive performance, and few studies have been conducted on techniques for self-position estimation and obstacle avoidance despite their indispensability in the autonomous navigation of fish-like robots. Therefore, this study aimed to explore a self-positioning estimation method for robotic fish in an environment where global positioning system (GPS) and pre-defined landmarks are not available. To this end, we first developed a fish-like robot that has a laterally flattened shape, which can mimic the swimming pattern of a fish moving forward by waving a tail fin. Next, we realized the function of simple obstacle avoidance using optical distance sensors for autonomous swimming from a practical perspective. Subsequently, we implemented a real-time swimming-path estimation using the posture derived from the inertial measurement unit (IMU) outputs and the swimming speed measured in advance. Furthermore, a swimming path correction method using a particle filter based on a pre-constructed magnetic map was investigated as an alternative to the GPS correction method. Experiments confirmed the accuracy of the swimming path estimation using the proposed method under various conditions, including obstacle avoidance.