The Marine Systems Group at the Australian Centre for Field Robotics conducts important research in sensing, navigation, visualisation and classification of marine habitats using autonomous underwater vehicles.
Our partners: Dr Neville Barrett (University of Tasmania), Dr Jacquomo Monk (University of Tasmania), Professor Gary Kendrick (University of Western Australia), Dr Renae Hovey (University of Western Australia), Dr Russ Babcock (CSIRO), Professor Peter Steinberg (UNSW), Dr Ezequiel Marzinelli (UNSW), Dr Tom Bridge (James Cook University), Dr Renata Ferrari (University of Sydney), Dr Will Figueira (University of Sydney), Professor Maria Byrne (University of Sydney)
Australia’s Integrated Marine Observing System (IMOS) Autonomous Underwater Vehicle (AUV) facility is providing precisely navigated time series measurements of water column parameters and benthic imagery using AUVs. AUV systems are effective for rapid and cost-effective high-resolution, accurately geo-referenced and targeted optical and acoustic imagery of the seafloor. Over the past decade we have conducted surveys of benthic reference sites around Australia. We have demonstrated how surveys of coral reefs using high-resolution imagery can track changes in reef environments and allow scientists to trace the recovery of these sensitive habitats from bleaching and tropical cyclones.
Our partners: Dr Donald G Dansereau (Stanford University)
The team is working on building a photometric model of the underwater image formation process and assisting with the design of novel underwater imaging systems. This uses rich information recorded across multiple apertures to achieve otherwise impossible tasks. By building noise filters tailored to the light field frequency, Dr Dansereau has demonstrated that a plenoptic camera can provide higher-quality imagery than a traditional camera in low light with murky water. We are now extending these models to include the design of imaging systems for challenging underwater conditions. As a result, robots will be able to operate in a wider range of environments.
Our external partners: Dr Brendan Foley from the Woods Hole Oceanographic Institution
A team from the Australian Centre for Field Robotics has participated in archaeological survey work in Greece. In 2014 and 2015 our team mapped a first century BC shipwreck off the coast of the Greek island of Antikythera using the AUV Sirius, equipped with a high-resolution stereo imaging system. We used multi-session simultaneous localisation and mapping techniques to stitch data from two dives into a single composite map of the site, and provided divers with a detailed 3D model of the site to facilitate excavation operations. Archaeological divers made use of the AUV-derived maps to plan and document their excavation work.
The archaeological team discovered a near-complete human skeleton on the wreck in late 2016. This astonishing find represents the oldest known human remains recovered from an underwater archaeological site and may provide scientists with opportunities to study the provenance of those who crewed the vessel.
Our experts: Dr Oscar Pizarro, Mr Christian Lees, Mr Lachlan Toohey
A centre team was involved in mapping the site of a submerged city at Port Royal, Jamaica. The work was part of a two-year project to map the city through the state-of-the-art development of marine robotic 3D mapping systems for underwater archaeology. The team employed a new multidimensional approach to surveying a shallow submerged site, using an autonomous surface vessel, a robotic stereo camera rig to conduct a 3D scuba survey of the site’s underwater buildings and a sector scan sonar survey, allowing the team to understand the extent of the site using acoustics.
The 3D maps reconstructed from these surveys will assist the Jamaican National Heritage Trust to apply for UNESCO world heritage status and advance the recognition of this spectacular underwater city.
Our partners: Dr Joshua Madin (Macquarie University), Dr Maria Dornelas (St Andrews University), Dr Ariell Friedman (Greybits)
We have been developing ways to survey shallow reef environments using diver-held underwater stereo imaging systems and small AUVs. A diver swims with a camera in a spiral pattern, unwinding a line around a drum to ensure even track spacing and a consistent survey footprint, something difficult to achieve without robotic assistance. They used this system to document the recovery of coral reefs following a cyclone that hit Lizard Island in the north of the Great Barrier Reef in 2014. These surveys have been repeated to document the impact of cyclones and subsequent recovery of these sensitive reefs.
Our experts: Mr Nader Boutros, Dr Oscar Pizarro
Nader Boutros has been developing ways to observe the spatial distribution of reef fish within their habitat. Observations were collected using a stereo imaging system to provide 3D localisation of their movement. The habitat was modelled using an image-based localisation and mapping algorithm. This technique facilitates the observation and modelling of interactions between fish and their habitat at a higher resolution than traditional sampling procedures.
Our industry partners: Dr Renata Ferrari Legorreta, Australian Institute for Marine Science
Renata Ferrari’s research produces data that can be used to help manage and conserve marine ecosystems. She is working with a multidisciplinary group to understand the benthic dynamics of marine habitats and their trajectory and change, with an emphasis on spatial and temporal patterns of biodiversity and processes.
Our experts: Dr Oscar Pizarro, Mr Christian Lees, Mr Andrew Durrant
Our external partners: Dr Blair Thornton (University of Southampton) and the Institute of Industrial Science at the University of Tokyo
Our deep-water stereo imaging system was deployed on a remotely operated vehicle to study artificial hydrothermal vents at 1km depth in Okinawa, Japan. This allowed the marine environment to be studied in greater detail than previously possible. Analysis of this data by scientists has shown a strong correlation between the complexity of the terrain, hydrothermal venting and biological community distribution. We have also examined how imaging data collected at multiple resolutions can facilitate surveys across large areas of seafloor, while allowing us to examine the taxonomic resolution at which species can be identified.
Andrew Durrant spent time with a team in Japan in mid-2014 surveying debris fields associated with the Fukushima tsunami. These studies documented the state of the marine environments and the impact of material washed offshore during the disaster.
A cooperative localisation project is being undertaken by Dr Lachlan Toohey on methods to allow asynchronous updates between robots over unreliable and low bandwidth communication channels. These new information fusion techniques allow for smaller summaries of new information to be created and transmitted, in contrast to many existing methods which have larger summaries that replicate previously transmitted information. This reduces bandwidth requirements and allows us to work with arbitrary communication networks.
Our experts: Mr Troy Wilson, Professor Stefan Williams
Troy Wilson has been investigating fine resolution sampling across space and time in estuaries using an autonomous surface vessel (ASV) to calibrate and validate long-term coarse resolution studies. The dynamic nature of tidal fluxes within estuaries is important due to the effect it has on mixing salinity, oxygen and nutrients between fresh water and salt water. The shallow nature of some parts of estuaries lends themselves to a small platform that can sample, model and plan without disturbing the environment.
By estimating models online, an autonomous moving platform can optimise sampling strategies to collect the most informative data. Troy has demonstrated how a small ASV with online planning capability can generate a map of a previously unknown seafloor contour. He is developing methods to allow such a vessel to model salinity levels around a moving tidal front.
Our experts: Mr Eric Ferguson, Mr Rishi Ramakrishnan
Eric Ferguson has been developing cost-effective approaches to remote monitoring of protected areas such as marine reserves and restricted naval waters using passive sonar to detect, classify, localise and track marine vessel activity. Eric has been working with Rishi Ramakrishnan to develop convolutional neural networks for the joint detection and ranging of broadband acoustic noise sources, in conjunction with a data augmentation approach in order to improve network performance in signal-to-noise ratio situations.
Our partners: The Schmidt Ocean Institute, Dr Chris Roman (University of Rhode Island), Dr Rich Camilli (Woods Hole Oceanographic Institution), Dr Angelos Mallios (University of Girona) and Dr Brian Williams (MIT)
The Australian Centre for Field Robotics has been working with The Schmidt Ocean Institute on tools and techniques to support the use of multiple robotic platforms to study sensitive tropical reef systems. We were able to demonstrate the simultaneous deployment of multiple autonomous underwater vehicle (AUV) systems, an unmanned surface vessel and an autonomous glider equipped with acoustic navigation and obstacle avoidance systems. These platforms were deployed to repeat surveys of coral reefs at Scott Reef on the North West Shelf of Australia.
Dr Mitch Bryson has been exploring ways to build environmental models using data collected by AUVs and low-cost aerial platforms such as kites. He has proposed automating colour correcting in underwater images to create evenly illuminated 3D structure-from-motion models. Mitch has developed an approach to colour correction that exploits the 3D structure of a scene created by structure-from-motion and photogrammetry techniques. These methods improve the consistency of photo-textured 3D models. He has also shown how ultra-high-resolution models of foreshore environments can be generated from the air using low-cost kites and imaging system.