Underwater Dynamic Visual Servoing for a Soft Robot Arm With Online Distortion Correction

Bioinspired soft robots have generated increasing attentions due to their optimal performance in specific environments stemming from their unique morphology and sensorimotor capabilities. This also raises new challenges in modeling and control due to their unconventional soft mechanism. In this paper, we aim to improve the controllability of an octopus tentacle-like soft robot arm operating in an underwater environment and to reproduce the perception-guided tracking performance of its biological counterpart when preying on and pursuing a target. To this end, we design an adaptive visual servoing controller for trajectory tracking tasks, taking into account system dynamics with environment interactions. Unlike on-the-ground visual servoing controller, the underwater environment exerts multirefraction effects on the image projection process, increasing difficulties in obtaining an accurate camera projection model. This paper thus proposes an adaptive method that can online correct image distortion caused by the deflection of light rays passing through different mediums to compensate for refraction effects. The adaptive algorithm can simultaneously estimate unknown camera parameters to completely avoid complex offline underwater camera calibration. The proposed adaptive controller validates the trajectory tracking performance experimentally and position and image velocity.