In this study, we describe a palm-sized robotic fish that can automatically track a goldfish. A robotic fish is suitable for ecological surveys because it is difficult to be noticed by aquatic animals. However, so far, there is no palm-sized robotic fish that can automatically track an aquatic animal. Automatic tracking by the robot is carried out by recognizing the goldfish with a camera and changing the swimming direction towards the goldfish. To follow the agile movement of the goldfish, the robot has a high turning ability with multiple joints. Finally, we have confirmed the robot can track a goldfish.

A robotic fish is ideal for an ecological survey of underwater life, because underwater animals cannot easily notice the robot. However, when a robotic fish swings its body to swim, it makes a rolling motion due to movements of the center of gravity and this rolling motion has a bad influence on the coordinate calculation of the tracking target, therefore it is necessary to correct the image. The robot has FPGA (Field Programmable Gate Array) and it is suitable for real-time image processing because FPGA has excellent performance with its parallel processing. In this study, we create a program to calculate the roll angle of the robot by using the gyro and acceleration sensor mounted on the robot and to correct the image by rotating it by that angle. Then, we conduct an evaluation of this rotation correction method with rotation tester that reproduces the rolling motion of the robotic fish and determine synchronization timing of image data and angle data because the CMOS camera used for the robot adopts rolling shutter method. Furthermore, we investigate the effect on coordinate calculation by the fish robot with created program. We have confirmed that this robot can perform real-time correction of rotated image and reduce the influence on the coordinate calculation of the target.