Project Overview

A reliable underwater acoustic communication is essential to transmit high-speed data for Australian offshore oil and gas industries, marine commercial operations, monitoring environmental pollution, collecting scientific data from seabed, communications between underwater robots and vehicles. However, the performance of underwater acoustic communication is restricted by time and frequency variation, reverberation caused by underwater obstacles, limited bandwidth and strong impulsive noise. For example, underwater vehicles are essential to inspect cracks on underwater pipes for oil and gas industries. Underwater vehicles capture videos or images for underwater pipes for inspection purposes. However, highly blurry or poor-quality videos can only be received under noisy environment. Therefore, developing accurate underwater communications is necessary.  

Conventional approaches in underwater communications only develop fixed models based on human knowledge or understanding which cannot fully cover the highly dynamic and complex characteristics of underwater communication channels; the performance of those classical models is poor, since understanding of underwater acoustic channels is insufficient.  

Recently our research group has implemented state-of-the-art deep learning for underwater communications; deep learning models underwater environment based on real data. Our preliminary study shows that state-of-the-art deep learning is able to achieve reasonable accuracies when the signal to noise ratio is high. To tackle real-time implementation and environment with low signal-to-noise ratio, more effective deep learning is essential. This project will develop more effective, accurate and simpler deep learning for underwater acoustic communications in real-time implementations.

Aims

We have implemented the state-of-art deep learning in simulated and indoor/outdoor environment. Reasonable results can be achieved in high signal-to-noise ratio environments; further research is required to improve deep learning in fast variation of underwater acoustic environment or low signal-to-noise environment. We will develop more effective deep learning to increase data communication speed and reduce error probabilities. The project aims to: 
- Developing underwater communication systems using deep learning which are well-performing to nonlinear channels. 
- Establishing a deep learning architecture which is optimal for underwater acoustic communication. 
- Developing simple deep learning models which have higher accuracies and require shorter execution time compared to conventional cumbersome deep learning. 
- Improving commonly-used underwater communication systems which have different performances under different channel conditions. 
- Providing high-speed underwater communication, for industrial applications in real time underwater communications.

Objectives

Overcoming the limitations of classical acoustic communication systems:  We will investigate joint transmitter and receiver designs for underwater acoustic communication systems. We will identify complicated system components which perform poorly by classical approaches. Those complicated components will be replaced with deep learning for achieving more accurate communication. 

 
Designing a simple deep learning architecture for real-time underwater acoustic communication: State-of-the-art deep learning is not practical to be embedded with limited resources in real-time, since deep learning is cumbersome and it involves a large number of computational operations. A network simplification approach will be designed to slim network sizes to suit real time implementations. 

Robust underwater acoustic communication for global ocean environments: Deep learning models perform robustly on certain environments since they are developed by data in low signal-to-noise ratio. To develop more robust models, we will collect real data from ocean and rivers. Deep learning models and architecture will be validated based on real data. More robust and accurate models will be developed for global ocean environments.

Significance 

This project focuses on developing digital technologies based on deep learning to increase transmission speeds and accuracies of acoustic communications in underwater environment. These digital technologies are crucial for Australian defence, offshore oil and gas industries, and ocean research. 
 

Undersea defence: Australian Defence Force will acquire nuclear-powered submarines under the enhanced trilateral security partnership, AUKUS. Underwater acoustic communication market growth will be strongly driven by spending on submarine building and undersea surveillance. The project outcome will enhance communication technologies which share or exchange signals or information between underwater and surface vessels in uncertain environments. 
 

Oil and gas industry: The proposed project will benefit oil and gas industries which are the industrial driving forces for Australia. The invented technologies will help autonomous underwater vehicles to look for submerged resources, and aid resource exploration and extraction. It helps oil and gas infrastructure defect detection and process reliability monitoring which are underwater. 
 

Scientific ocean research: Ocean has rich sources for energy and minerals. Ocean exploration and data collection is increasingly important to help understand planetary-scale processes including tectonics and marine hazards; it helps to explore energy, mineral and biological resources. The underwater acoustic communication technologies will help.

The school is focusing on research in AI/machine learning and signal processing which are the research areas in this proposed project. We have equipment ($30K) for collecting underwater communication data from the ocean or the laboratory. 
A laboratory equipped with a water tank is available to simulate the environment of underwater acoustic communications. Evaluation of model performance can be conducted based on the data collected through the water tank.  

We have the GPU machines ($14k) to develop deep neural networks for underwater communications. 
The supervisors are experts in deep learning, machine learning and underwater acoustic communications. Dr. Chan (total citations 5282; h-index 38) has good track record in deep learning and machine learning technologies for various applications including underwater acoustic communications. Prof. Rong (total citations 4744; h-index 37) has good track record in underwater acoustic communications. The supervisors are working with a PhD student and a research fellow who have collected underwater data and preliminary algorithms. With their guidance and supervision, project aims and objectives are expected to be achieved by the HDR candidate. Also, the supervisor team has external funds from the Defence Science and Technology Group to cover the project expenses of fieldworks, conferences, travels, publications etc.

An internship may be available for this project. We have strong ongoing collaborations with WA companies specialising in underwater communications, which provide internship opportunity for HDR students. We also have experience to supervise ICP internship students who do AI and deep learning topics.

• Future Students

• Faculty of Science & Engineering
  • Science courses
  • Engineering courses

• Higher Degree by Research

• Australian Citizen
• Australian Permanent Resident
• New Zealand Citizen
• Permanent Humanitarian Visa

• Merit Based

The annual scholarship package, covering both stipend and tuition fees, amounts to approximately $70,000 per year.

In 2024, the RTP stipend scholarship offers $35,000 per annum for a duration of up to three years. Exceptional progress and adherence to timelines may qualify students for a six-month completion scholarship.

Selection for these scholarships involves a competitive process, with shortlisted applicants notified of outcomes by November 2024.

1

All applicable HDR courses.

Outstanding Computer Science or Engineering student 
Knowledge of data science and machine learning 
Knowledge with Python or Matlab.

 

Please send your CV, academic transcripts and brief rationale why you want to join this research project via the HDR Expression of Interest form to the project lead researcher, listed below. 

You must be enrolled in a Higher Degree by Research Course at Curtin University by March 2025.

Project Lead: Dr Kit Yan Chan