Skip to end of metadata
Go to start of metadata

 

Chief Investigators

Dr Matthew Dunbabin

Prof Peter Corke

 

Research Fellows

Dr Feras Dayoub

 

Enquires

Forward any enquiries to m.dunbabin@qut.edu.au.

 

The COTSbot (Crown-Of-Thorns Starfish robot) aims to be a revolutionary advancement in robotic environmental monitoring and management, specifically to increase the efficiency of Crown-Of-Thorns Starfish (COTS) eradication. Integrating state-of-the-art robotic vision and classification algorithms with experience in, and technologies for, shallow coastal water robotic monitoring, COTSbot also aims to provide a flexible tool that empowers a range of stakeholders to scale current eradication programs and protection of reefs such as the Great Barrier Reef.

 

Background

Crown-Of-Thorns Starfish (Acanthaster planci) are described as one of the most significant threats to the Great Barrier Reef. Since the 1960's, land-based nutrient runoff has accelerated outbreaks of COTS which are destroying large areas of reef. With few natural predators, traditional control of COTS required manually injecting the starfish in excess of 10 times with a biological agent. In 2014, a new agent was released, which was developed by the James Cook University (JCU) requiring only one injection per starfish. This advancement provided the stimulus for us to revisit automated (robotic) COTS population control and monitoring.

The intention of COTSbot is to provide a proof-of-concept robotic system that consolidates prior and ongoing research into image-based COTS detection, robotic vision, manipulator control, and shallow water Autonomous Underwater Vehicle (AUV) design, navigation and control, to directly facilitate COTS reduction. This multi-stage project will validate and demonstrate AUV performance to stakeholders and ensure the system components are a useful and flexible enabling foundation technology for environmental monitoring beyond the problem of COTS control. 

The COTSbot Platform 

The COTSbot is an Autonomous Underwater Vehicle (AUV). This class of robot is designed to operate without a tether and execute missions with minimal human interaction once deployed. The primary task of the COTSbot AUV is to autonomously navigate within complex reef environments and automatically detect COTS on the coral and administer an injection. The following summarises the robotic and operational requirements of the AUV and the resulting prototype design. The fundamental robotic capabilities required by the COTSBot AUV prototype are:

  • Real-time and on-board automated image-based detection of COTS.
  • Real-time, autonomous injection of bile salts into detected COTS.
  • Autonomous altitude control within a shallow reef environment (< 30m).
  • Autonomous transects for coverage survey.

Based on the desired goal of COTS eradication, the physical and operational design criteria are proposed for the AUV are:

  • Maximum in-air weight: 26-30 kg.
  • Speed range: survey 0.5 ms-1, max >2 ms-1.
  • Endurance: >6 hours.
  • Maximum depth rating: 100 m, typical operation <30 m.
  • Ability to station keep (hover) briefly for injection of COTS.

The primary design considerations related to the navigation system, vision system, computing system, propulsion, power, and mechanical design and physical layout. The resulting COTSbot prototype is illustrated below.

 

Vision System

A fundamental component of COTSbot is its ability to automatically and robustly detect Crown-Of-Thorns Starfish in complex coral reef environments. In this project, a new state-of-the-art vision and classification system has been developed to run on-board the AUV.

Previous research on image-based COTS detection has increasingly improved the detection accuracy of the system and its ability to run on-board modern lower power computers which can be installed within the space and power constraints of the underwater robot (see publications below). The current COTSbot system is a significant further advancement that exploits state-of-the-art techniques in machine learning to achieve a detection performance of well over 99% whilst being able to operate real-time on-board the AUV. The videos below illustrate the detection performance on image sequences collected on the Great Barrier Reef.

 

 

 

Related publications

  1. Dayoub, F., Dunbabin, M. and Corke, P. (2015). Robotic detection and tracking of Crown-of-thorns starfish. To Appear In Proc. 2015 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS),  pdf. * 
  2. Clements, R., Dunbabin, M. and Wyeth, G. (2005). Towards robust image detection of crown-of-thorns starfish for autonomous population monitoring. In Proc. Australasian Conference on Robotics and Automation 2005, pdf. 

Other Links

For information on robotics related research at the Queensland University of Technology please visit the following links.

robotics@QUT home page (here).

robotics@QUT YouTube site (here).

Australian Centre for Robotic Vision (here).

International RobotX Maritime Challenge team site (here) .

* (This paper explains the proposed framework used for tracking and detection. Since this work was completed we have improved the precision rate greatly.)

  • No labels