Development of Visual Servoing-Based Autonomous Docking Capabilities in a Heterogeneous Swarm of Marine Robots

Nikola Mišković,Anja Babić,Filip Mandić

doi:10.3390/app10207124

Abstract

This paper describes the design, development, and testing of both hardware and software for a visual servoing-based system that enables agents within a heterogeneous marine robotic swarm to share energy. The goal of this system is prolonging the active operational time of the swarm as a whole, allowing it to perform long-term environmental monitoring and data collection missions. The implementation presented in the paper features an over-actuated autonomous surface platform docking up to four floating sensor nodes at a time and replenishing their batteries using wireless inductive charging. Mechanical solutions for each robot segment related to the docking procedure are presented, along with pertinent high-level and low-level control structures. A node featuring an extended Kalman filter and additional heuristics is used to fuse measurements from a neural network trained for object detection and a hue thresholding image processing algorithm, in order to track the docking target and achieve visual servoing. Finally, experimental results in both a controlled environment and challenging conditions on-site are presented. Once deployed, the developed system successfully enables the approach and docking of the designated target, showing its feasibility in different real-life conditions.

Highlights

The EU Horizon 2020 FET project subCULTron aims to develop a heterogeneous swarm of marine robots and deploy it on an unsupervised mission of long-term marine monitoring and exploration in the challenging environment of the lagoon surrounding Venice, Italy [1,2]
The mechanical systems, hardware, and software necessary for autonomous docking have all gone through several iterations of development, testing, and redesign, though the basic concept has remained the same throughout: the aPad seeks the recognizable top cap of a floating aMussel using a visual sensor, approaches it using a control algorithm-based around visual servoing, moving so the aMussel docking section is aligned with the aPad docking mechanism, activates the docking mechanism to secure the aMussel in place so wireless charging using inductive coils can take place
Previous work can be found in [4] in which the initial indoor experiments which served as proof of concept for the wireless charging are described, as well as some initial outdoor experiments aimed at testing the capabilities of the aMussel and aPad positioning systems

Summary

Introduction

The EU Horizon 2020 FET project subCULTron aims to develop a heterogeneous swarm of marine robots and deploy it on an unsupervised mission of long-term marine monitoring and exploration in the challenging environment of the lagoon surrounding Venice, Italy [1,2]. The three agent types are as follows: five over-actuated autonomous surface vehicles (ASV) called aPads or artificial lily pads; a small swarm of highly mobile aFish or artificial fish; and 120 underwater sensor nodes with limited mobility called aMussels, or artificial mussels. This paper focuses on an algorithm developed primarily for interaction between the aPad and aMussel agent types. Equipped with mechanical docking stations and inductive charging coils, one of the aPad’s main roles and most important abilities is to transport and wirelessly charge up to four other swarm agents at a time. Two antennas, and an acoustic modem, it plays an important role in the swarm’s communication structures and possible underwater-swarm-to-surface-observer information transfer

Methods

Results

Conclusion