Investigation of biomimetic systems and their applications in robotic solutions in fluids: using deep learning and vision-based control

Impressive robotic solutions with astonishing capabilities have been designed to inherit certain propensities of living creatures and mimic some of their abilities in order to perform real-life tasks. In the present work, two bio-inspired systems are developed to investigate certain characteristics of biological hair-like structures and employ their behaviours in robotic and sensory solutions in fluids. The first is an intelligent sensor that is developed using deep learning to detect the position of underwater wake-generating objects inspired by the seal's ability to track its prey by sensing the surrounding fluid motion using its facial hairs (i.e., whiskers). The produced sensor provides a safe, passive and lifelike way of underwater sensing which can be utilized in robotic applications for underwater navigation in dark or cloudy environments and in situations that require stealth. The second is a mechatronic system that is designed and implemented to resemble the self-organization of biological cilia in an enlarged model and to generate flow propulsion at low-Reynolds regimes using the metachronal coordination of rotational oscillators. Using two dimensional flexible flat plates (i.e., flaplets) as the oscillating (beating) elements, a metachronal-wave pattern is experimentally proven to spontaneously emerge due to the hydrodynamic interaction among the oscillators. A mathematical model of the physical system is then developed for a better understanding of the coordination collective effects and for analysing its stability. The model can track the emerged coordination over long periods and estimate the net propulsive force acting on the physical model. The developed system is able to produce effective propulsion, that can be utilized in robotic applications, despite the time-symmetric beating profiles and the single degree-of-freedom actuation of the individual oscillators