Our projects

Our project aims to enhance disaster response simulations using advanced graph-based algorithms and machine learning. Led by CIs Dr. Shan (Emma) Xue, Prof. Amin Beheshti, Prof. Mohsen Asadnia, A/Prof Jia Wu, and GRA students Ms. Qihua Lyu and Mr. Yifei Han, the project focuses on improving the scalability, flexibility, and adaptability of simulated agents to better coordinate rescue teams in post-earthquake scenarios. Key outcomes include robust pathfinding algorithms, an adaptive learning framework for agents, enhanced communication models, and a user-friendly interface for simulation control and data analysis, contributing to more effective real-world disaster response strategies. These advancements will enable more efficient coordination and resource allocation in disaster situations, ultimately improving the effectiveness of emergency response efforts.

Research team: Dr Shan (Emma) Xue; Prof Amin Beheshti; Prof Mohsen Asadnia; A/Prof Jia Wu; Ms Qihua Lyu; Mr Yifei Han

Forgeries pollute data sets of all kinds and compromise their value. Forgers often evade detection by varying their practice as they become more fluent in the use of particular strategies to advertise the authenticity of their wares. Understanding how forgers develop and deploy their skillsets over time in a responsive way is critical to anticipating security risks. Our pilot study applies Change Point Analysis to detect changes in the handwriting style used by the 19th century forger, Constantine Simonides, on forged ancient papyrus manuscripts. Through this case study we will reconstruct a model of how individual competency in deception was developed and deployed.

Research team: Dr Rachel Yuen-Collingridge; A/Prof Georgy Sofronov

This project utilises Sentinel-2's multi-spectral satellite imagery to generate an upscaled image using multi-frame super-resolution (MFSR) method. The single-frame method uses only one low-resolution image to generate the high-resolution image. In contrast, the multi-frame method uses multiple low-resolution images to reconstruct hidden high-resolution details not present in the low-resolution image. This pilot project tested residual attention deep neural networks (RAMS) and derived the MFSR Sentinel-2 true colour imagery from the original nine images at 10m resolution, upscaling to an image at 3.3m resolution. Future development involves training the model to correctly upscale images using other spectral bands and sensors.

Figure: True colour satellite of Macquarie University collected by Sentinel-2. (Left) original image at 10m resolution, and (right) upscaled image at the resolution 3 times higher.

Research team: Dr Michael Chang; Prof Hanlin Shang; Prof Benoit Liquet-Weiland; Dr Teo Nguyen

Studies show that seniors with sensory loss are severely disadvantaged when accessing and using online information and services. We seek to close this growing digital divide by developing a novel conceptual framework merging theory of change and social justice theory to underpin the design and piloting of a bespoke digital literacy program for seniors with sensory loss. This research leverages the capabilities of 7 partner organisations: Blind Citizens Australia; Hearing Matters Australia; Meals on Wheels NSW; YourLink; Soundfair; Collective Leisure; and Communiteer, who can offer valuable insights and serve as effective frontline digital mentors to enhance the digital inclusion of seniors with sensory loss through skill-building and connectivity.

Research team: Prof Bamini Gopinath; Prof Lisa Keay (UNSW); A/Prof Melanie Ferguson (Curtin University); Dr Chyrisse Heine (Federation University); Prof Niloufer Selvadurai; A/Prof Kompal Sinha; Dr Diana Tang; Dr Sheela Kumaran (UNSW)

This project aims to enhance the quality of science learning by developing AI-generated feedback on students' learning progress. It focuses on utilising advanced AI techniques to analyse student understanding, identify knowledge gaps, and provide resources for improving comprehension. Specifically, the project involves creating an AI-generated feedback platform for pre-service teachers and tertiary students, as well as developing and implementing an AI-powered chatbot for elementary schools.

Research team: Dr Hye-Eun Chu

Superannuation industries carry significant responsibility for the financial well-being of Australia. Managing optimal investment strategy in the superannuation accumulation phase depends on optimally constructing a portfolio of assets. This project aims to develop new theories, methodologies, and algorithms that account for distributional uncertainties in high-dimensional datasets to improve the optimality of portfolio construction. Our research will build the industry's capacity to use these new methodologies, leading to a distributionally robust mean-variance portfolio that can achieve optimal investment returns for everyday Australians.

Research team: Prof Hanlin Shang; Ruike Wu (Xiamen University); Yanrong Yang (ANU)

Learn more about this project here.

The Perachora Peninsula Archaeological Project is conducting intensive surface survey across a town associated with the 8th-2nd century BCE Sanctuary of Hera Akraia located opposite Corinth. The collected artefact samples are indicative of land use and the chronological extent of the habitation, but projecting population densities is problematic. With the collaboration of our Data Horizons colleagues, the Perachora Peninsula Project aims to apply new interpretive tools to our archaeological data to arrive at a more satisfying projection of the population that inhabited the town – a method which could then be applied to other regions that are also being investigated with intensive surface survey.

Photograph by Dr Susan Lupack

Research team: Dr Susan Lupack

Poor understanding of speech in noise continues to be a perplexing problem in hearing science. Approximately one in ten people with speech-in-noise hearing difficulties cannot be helped
because the underlying cause of their hearing problem remains a mystery. Data Horizons will develop a new data-centric approach to brain network analysis: connectome-based predictive modelling, to predict individual speech in noise performance from brain connectivity. The structure of the human connectome could inform artificial agents that are able to mimic the way a human interacts with the world, both in health and disease.

Research team: Prof Paul Sowman; A/Prof Jia Wu; Prof Jian Yang

Australia’s life insurance, superannuation and pension funds industries carry significant responsibility for the financial wellbeing of Australians. Managing this responsibility and financial risk depends on accurately pricing consumers’ insurance premiums. This project will develop new theories, methodologies and algorithms that account for complexities in merged big datasets to improve the accuracy of predictions. Our research will build industry’s capacity to use these new methodologies leading to improvements in mortality forecasts and pricing of life insurance premiums for everyday Australians, as well as stronger financial risk management among some of Australia’s most critical financial industries.

Research team: Prof Hanlin Shang; Mr Sizhe Chen; Dr Yang Yang (UON)

Learn more about this project here.