Human mobility analytics is essential for applications such as intelligent transportation systems, network optimisation, urban planning, epidemic modelling, and next-generation communication networks. Advances in mobility sensing technologies, including mobile devices, GPS, and IoT sensors, have enabled the collection of large-scale, fine-grained mobility data, improving modelling and prediction capabilities. However, several challenges limit the effective use of these data.
Privacy concerns restrict data sharing because mobility traces can reveal sensitive personal information, including daily routines and frequently visited locations. In addition, many available datasets are incomplete, biased, or geographically limited, reducing their generalisability. Human mobility patterns are also highly complex, dynamic, and multi-scale, making them difficult to model accurately. Synthetic data generation has emerged as a promising solution by producing data that preserve the statistical and behavioural characteristics of real mobility traces while protecting privacy. Although deep generative models have demonstrated significant potential, challenges remain in generating realistic out-of-distribution (OOD) or unseen mobility patterns, preserving semantic and contextual relationships, and balancing data utility with privacy protection.
This project proposes a comprehensive framework for generating realistic, privacy-preserving, and controllable synthetic human mobility data. A key focus is the controlled generation of unseen or out-of-distribution (OOD) or unseen mobility behaviours by learning latent mobility patterns from both population-level and individual-level data, enabling robust modelling under evolving real-world conditions. For example, the framework could generate synthetic commuting patterns for weekends in City 1 by leveraging a user’s observed weekend mobility behaviours in City 2 together with population-level mobility patterns in City 1. In this scenario, even before the user visits City 1, likely destinations, visitation frequencies, and dwell times can be estimated and synthetically generated. Such capabilities could support proactive applications, including traffic forecasting, intelligent transportation management, network bandwidth provisioning, and personalised location-based services.
The project will investigate the integration of foundation AI models to analyse mobility behaviour at both population and individual levels with privacy preservation and extract transferable mobility knowledge. These insights will be combined with privacy-preserving learning mechanisms and quantum-enhanced generative techniques for controlled generation of synthetic OOD or unseen mobility patterns. The resulting framework will be implemented as a modular, user-friendly tool designed to support both academic research and industrial applications.
Aim
The primary aim of this project is to design and develop a high-fidelity, privacy-preserving synthetic human mobility data generation framework capable of controlled generation and modelling of complex mobility behaviours, with a particular focus on out-of-distribution patterns, using advanced generative AI techniques, including foundation models and quantum-enhanced approaches. The project will incorporate privacy-preserving mechanisms to protect individual mobility data while enabling effective and utility-preserving synthetic data generation.
Objectives
- To analyse and characterise real-world human mobility datasets to identify key spatial, temporal, and contextual patterns underlying user behaviour, capturing both individual-level and population-level dynamics using foundation AI models.
- To develop methods for inferring properties of unseen or OOD mobility patterns and trajectories by leveraging knowledge learned from observed population level distributions and limited user mobility data of known patterns.
- To explore and integrate quantum generative models for generating complex mobility patterns, including unseen and OOD/unseen behaviours, incorporating insights from O1 and O2 for controlled generation and validation.
- To investigate potential privacy leakage in controlled mobility data generation and integrate state-of-the-art privacy-preserving mechanisms into the generative models, along with evaluation of the privacy–utility trade-off.
- To develop a user-friendly, modular software platform that integrates all components of the framework, ensuring extensibility, robustness, and ease of adoption for future research and applications.
Significance
This project has significant impact across scientific, technical, economic, and societal dimensions.
From a scientific and technical perspective, it advances human mobility analytics, generative AI, and privacy-preserving data synthesis by enabling realistic modelling of complex and out-of-distribution mobility patterns. It also explores quantum-enhanced generative approaches and delivers a modular, reproducible framework that can be extended for mobility-aware optimisation in both research and industry applications, supporting future integration with large-scale intelligent systems.
From an economic perspective, the framework reduces reliance on costly and restricted real-world mobility data collection, storage, and sharing, while easing regulatory constraints. More importantly, it enables improved prediction of unseen mobility behaviours, supporting more efficient decision-making in traffic management, telecommunications, and urban systems, with direct benefits for productivity and infrastructure optimisation.
From a societal perspective, the project enables more effective and personalised mobility-aware services while strengthening privacy protection through realistic synthetic data generation. By reducing exposure of sensitive human movement data, it promotes ethical and responsible AI use and builds greater trust in data-driven systems for long-term societal benefit.
Ideal Candidate
We are seeking a highly motivated PhD candidate with strong analytical, problem-solving, and research skills to join this project on AI-driven human mobility analytics and synthetic data generation.
The ideal candidate should have a background in Computer Science, Data Science, AI/ML or a related field, with strong quantitative and programming skills (Python preferred). Familiarity with machine learning, deep learning, generative AI, and spatiotemporal data analysis is highly desirable. Experience with privacy-preserving learning, foundation models, or quantum computing will be an advantage.
Candidates should be self-motivated, well-organised, and able to work both independently and collaboratively. Applicants must be eligible to enrol in a PhD program at Curtin University.
This project is open to International and Domestic applicants.
Scholarship
If you are identified as the preferred candidate for this project, you may be considered for an RTP scholarship.
Enquires and How to Apply
For enquires about this opportunity contact Dr Chamara Manoj Madarasingha Kattadige at C.Kattadige@curtin.edu.au
To formally apply submit an Expression of Interest to Dr Chamara Manoj Madarasingha Kattadige during the Central Scholarship round (July 1st – July 31st 2026)