Multi-agent Reinforcement Learning for Robotized Coral Reef Sample Collection

TL;DR

This paper develops an autonomous underwater robotic system for coral sampling using reinforcement learning, digital twins, and real-time motion capture, enabling effective simulation-to-real transfer for reef conservation tasks.

Contribution

It introduces a novel combination of game engine simulation, deep RL, and underwater motion capture for zero-shot sim-to-real transfer in coral reef sampling robots.

Findings

Successful deployment of RL-trained controller in physical experiments

Effective synchronization between simulation and real-world hardware

Demonstrated potential for autonomous reef sampling applications

Abstract

This paper presents a reinforcement learning (RL) environment for developing an autonomous underwater robotic coral sampling agent, a crucial coral reef conservation and research task. Using software-in-the-loop (SIL) and hardware-in-the-loop (HIL), an RL-trained artificial intelligence (AI) controller is developed using a digital twin (DT) in simulation and subsequently verified in physical experiments. An underwater motion capture (MOCAP) system provides real-time 3D position and orientation feedback during verification testing for precise synchronization between the digital and physical domains. A key novelty of this approach is the combined use of a general-purpose game engine for simulation, deep RL, and real-time underwater motion capture for an effective zero-shot sim-to-real strategy.