Multi-Robot Visual Control of Autonomous Soft Robotic Fish

Abstract

The coordination and control of autonomous underwater vehicle (AUV) fleets in ocean exploration is a widely researched topic with much groundwork for traditional AUVs. Depending on the mission, AUV fleets can relax mission constraints on individual vehicles and improve a number of performance objectives (e.g. duration, sampling rate, area coverage). As missions begin to require navigation within more confined areas such as caves and coral reefs, however, safe interaction with such environments becomes more difficult for typical rigid AUVs and more feasible for soft, compliant underwater robots that can adaptively deform to their surroundings. Moreover, soft underwater robots show great promise as biomimetic vehicles that can take inspiration from nature’s swimmers and help answer questions about their behavior, for instance about the schooling capabilities observed in many fish species. Unfortunately, few fully autonomous, self-contained underwater soft robots have been developed, let alone fleets of such robots. To address this, we present a milestone towards formation control of a fully autonomous, multi-soft robotic fleet inspired by fish schooling. We present a vision-based, leader-follower formation strategy using an untethered soft robotic fish (SoFi) platform that enables one SoFi robot to pursue another via a visual servoing behavior. Our system demonstrates basic formation control of a pair of fully autonomous, self-contained soft robotic fish without external input.