A novel active tracking system for robotic fish based on cascade control structure

Abstract

This paper presents a novel active tracking control approach for a self-propelled robotic fish with onboard vision system. Considering the image instability caused by the swaying of the fish head, a camera stabilizer for visual system is designed to obtain stable images. Meanwhile, a feedback-feedforward controller is developed to maintain the attitude of camera to the inertial frame based on two Inertial Measurement Units (IMU) separately fixed on the camera and the fish body. In order to effectively track a target object, the Kernelized Correlation Filters (KCF) is adopted and a relevant active tracking controller is also designed. With the feedback-feedforward controller as an inner loop and the active tracking controller as an outer loop, a cascade control system is formed. Finally, the simulation and experiment results both verify the effectiveness of the mechanism design for the camera stabilizer and the corresponding cascade control approach for the active tracking system.