TIME
SESSION
LOCATION
8:00 – 9:00
Registration and morning networking
Great Hall Foyer
9:00 – 10:30
Opening
Plenary 3: Industry–Academic Dialogue on Future Transport Research Priorities
Panel discussion moderated by Jonathan Busch, (SCT Consulting)
UTS Great Hall
10:30 – 10:50
Morning Tea
Great Hall Foyer
11:00 – 12:30
Paper presentation period 5
Paper 100
Traffic Detector Structural Data Quality Diagnosis via Weakly Supervised Fundamental Diagram Representation Learning
Room Location: B11.B2.101
View Authors
- Xiexin Zou (The Hong Kong Polytechnic University)
- Edward Chung (The Hong Kong Polytechnic University)
View Abstract
| Reliable traffic data are essential for intelligent transportation systems, yet operational traffic detectors are vulnerable to camera obstruction, calibration drift, communication interruption, and long-term sensing degradation. This paper addresses traffic detector structural data quality diagnosis by examining whether daily observations remain consistent with the historical fundamental diagram (FD) signature of each detector. A weakly supervised framework is proposed that avoids explicit anomaly labels by learning detector identity from historical detector-day observations. The model adopts a dual-branch architecture to jointly encode daily temporal traffic dynamics and detector-specific FD structural patterns. During inference, an anomaly scoring mechanism is introduced to detect structural abnormalities. The framework was evaluated using data from 756 traffic detectors in Hong Kong. On real-world data, the model achieved 98% detector-identification accuracy, confirming the existence of stable detector-specific signatures. Real-world structural diagnosis further revealed persistent anomalies concentrated in a limited subset of detectors. To enable controlled validation, an injected-anomaly benchmark was constructed from real traffic data. The proposed anomaly scoring mechanism achieved an F1-score of 0.96 and showed strong sensitivity to scale drift, sustained undercounting, variable dropout, variable desynchronisation, and FD shape deformation. The results indicate that some anomalies preserve plausible traffic values while violating detector-specific structural consistency. The proposed framework provides an interpretable and scalable basis for automated detector monitoring, maintenance prioritisation, and reliability-aware traffic analytics. |
Paper 118
From automated vehicle trials to freight deployment in Western Australia
Room Location: B11.B2.101
View Authors
- Brett Smith (The University of Western Australia)
- Doina Olaru (The University of Western Australia)
View Abstract
| Australia is moving from trial-based governance of automated vehicles toward a permanent national safety framework through the proposed Automated Vehicle Safety Law (AVSL). This paper provides policy advice for Western Australia on how to prepare for automated freight trials and early deployment under this emerging framework. It argues that the AVSL will strengthen the national regulator role by clarifying certification, in-service safety duties and the accountability of the Automated Driving System Entity, but that Western Australia’s deployment readiness will also depend on state-level facilitation and selective participation. Drawing on Australian automated vehicle policy documents, the 2024-27 National Connected and Automated Vehicle Action Plan, Austroads physical and digital readiness work and wider AV governance literature, the paper identifies the practical implications for freight corridor selection, trial approvals, road-manager data, telecommunications, institutional capability and staged deployment. The paper considers trials as learning systems that build evidence for AVSL-aligned deployment, particularly in mining-related logistics, hub-to-hub freight and predictable trunk corridors. The paper recommends a Western Australian automated freight deployment-readiness framework that integrates regulatory, physical, digital, institutional, freight-system and workforce dimensions. |
Paper 142
Mobility-as-a-Service preferences and willingness to pay under flat-fare public transport: Evidence from Southeast Queensland, Australia
Room Location: B11.B2.101
View Authors
- Aitan Medeiros Militao (Queensland University of Technology)
- Prithvi Beeramoole (Queensland University of Technology)
- Zeke Ahern (Queensland University of Technology)
- Alexander Paz (Queensland University of Technology)
View Abstract
| Mobility-as-a-Service (MaaS) has been explored as a means of integrating multimodal transport services through a single subscription, with the broader objective of reducing private car use and supporting more sustainable urban mobility. Empirical evidence on consumer preferences for MaaS bundles, however, remains concentrated in a small number of jurisdictions, and key questions about how those preferences interact with local public transport pricing regimes are largely unanswered. This paper provides empirical evidence on MaaS preferences and willingness to pay (WTP) in Brisbane and South East Queensland (SEQ), Australia, where a network-wide 50-cent flat fare for public transport was introduced in August 2024. A stated-preference survey was conducted across SEQ using a "design-it-yourself" bundle approach, in which each respondent constructed a personalised monthly MaaS bundle through an adaptive four-step process. A mixed multinomial logit model is estimated, and a simulation analysis is performed using a synthetic population to derive spatial adoption indicators. Bundle uptake is positively associated with car-sharing days, daily e-scooter minutes, and taxi/rideshare credit. Under our design and price settings, unlimited public transport inclusion is not a significant driver of bundle uptake, suggesting that where the marginal cost of public transport is already low, its role within MaaS is better understood as an integrated foundation of the service rather than a priced differentiator. Simulation results show that adoption is spatially concentrated, with predicted uptake highest in urban catchments with higher residential density and more mixed land use and consistently sensitive to the implied subscription fee across all bundle scenarios tested, providing both geographic and pricing reference points for the design and targeting of early-stage MaaS pilots in SEQ. |
Paper 154
Benchmarking deep reinforcement learning, metaheuristics, and exact solvers for urban tourist bus routing
Room Location: B11.B2.101
View Authors
- Bao Le (RMIT University)
- Sara Moridpour (RMIT University)
- Minh Dinh (RMIT University)
- Bao Nguyen (RMIT University)
View Abstract
| Urban tourist bus routing is a multi-objective optimisation problem in Intelligent Transportation Systems (ITS) that requires balancing scenic rewards with strict operational and temporal constraints. Traditional approaches, including exact solvers and advanced metaheuristics like NSGA-II, face significant computational latency, often requiring minutes to hours to explore large combinatorial spaces, making them less suitable for rapid itinerary adjustments. To address this limitation, this paper presents a detailed benchmarking study that evaluates a Double Deep Q-Network (Double DQN) framework against mathematically rigorous baselines. The models are benchmarked using the real-world, heavy-vehicle-filtered urban road network of Da Nang, Vietnam, evaluating trade-offs between accumulated scenic reward and operational costs. Empirical results demonstrate that Double DQN achieves competitive route quality, capturing an average of 95.5% of the best-known bound produced by Google OR-Tools, with an average gap of 4.5% relative to these bounds. Also, the proposed framework shifts the computational burden offline, reducing inference latency to approximately 25.4 milliseconds, an acceleration by several orders of magnitude compared to the hours required by exact solvers and the ~18.5 minutes needed for NSGA-II. These findings indicate that the proposed framework successfully delivers logical, spatially coherent itineraries, enabling competitive multi-objective routing with millisecond-level latency, offering a potential foundation for future dynamic routing models and what-if scenario analysis. |
Paper 87
Designing hybrid micromobility systems with integrated charging and rebalancing
Room Location: B11.B2.103
View Authors
- Reza Saadati (The University of Queensland)
- Mehmet Yildirimoglu (The University of Queensland)
- Jiwon Kim (The University of Queensland)
View Abstract
| The rapid growth of shared micromobility services has introduced new challenges in the strategic design and operation of integrated systems, particularly when combining station-based (SB) and free-floating (FF) systems. Existing studies typically address facility location, charging infrastructure, or rebalancing operations separately, often overlooking the interactions between these components over time. This paper proposes a strategic design framework for hybrid shared micromobility systems that jointly determines the location and capacity of stations and FF zones, while incorporating charging infrastructure and rebalancing operations. The model explicitly captures user behaviour through a logit-based mode choice formulation, where users choose between micromobility and an alternative mode based on generalised travel costs, including walking and riding components. A key contribution of this work is the integration of an energy layer that tracks the aggregate state of charge (SoC) of micromobility devices over time. This enables the endogenous estimation of charging demand at stations and battery management requirements in FF zones. Rebalancing is modelled at a strategic level to account for spatial and temporal imbalances in device availability without introducing full vehicle routing complexity. The resulting formulation is a time-dependent mixed-integer optimisation model that captures the interactions between infrastructure design, user demand, and energy constraints. The model is implemented on a test network to demonstrate its applicability and to provide insights into system performance under different design scenarios. The results highlight the importance of jointly optimising infrastructure, charging, and rebalancing decisions, showing that integrated design can significantly improve system efficiency and service coverage. |
Paper 102
Characteristics of e-bike injuries: Findings from American roadways
Room Location: B11.B2.103
View Authors
- Kevin Fang (Sonoma State University)
- Asha Weinstein Agrawal (Mineta Transportation Institute)
- Amelia Le (Mineta Transportation Institute)
View Abstract
| Electric bicycles have surged in popularity in Australasia and the world in recent years. However, the rise of electric bicycles has been accompanied by concerns about safety, attracting news media articles, editorials, and attention from legislators such as a parliamentary inquiry in Queensland released earlier this year. This paper explores what might be learned about electric bicycle safety from a look at injuries across the ocean in the United States. In particular, we look at electric bicycle-related injuries from 2020-2024 documented in the National Electronic Injury Surveillance System (NEISS), a database tracking emergency room visits at 96 hospitals across the U.S. Overall, we found subtle differences in injury outcomes between electric bicycles and conventional bicycles. Electric bicycle patients (16%) had a higher hospitalisation rate versus conventional bicycles (13%). Additionally, electric bicycle patients (23%) had a higher rate of head injuries than conventional bicycles (19%). That said, when narrowing the examination to only injuries which took place on streets, the gap in hospitalisation rates and head injury rates between the two devices disappeared. Much of the discourse on electric bicycle safety in both Australasia and the U.S. have revolved around risks to minors. However, NEISS data show that electric bicycle patients in the U.S. skew substantially older than conventional bicycle patients. Some of this is presumably due to children learning to ride conventional bicycles at younger ages than electric bicycles. However, it also highlights the importance of looking at electric bicycle safety in adults, particularly older adults who had the highest hospitalisation rates observed. Most injured electric bicycle riders were involved in solo crashes, though nearly a third were involved in vehicle collisions, highlighting the challenge of vehicular conflicts for both human-powered and motorized bicycles. 4% of electric bicycle patients were bystanders involved in a crash with an electric bicycle. While a low single-digit percentage, this was higher than the 1% bystander proportion among those injured with conventional bicycles. |
Paper 127
Using Intervention Mapping to develop behavioural interventions supporting young women’s cycling uptake
Room Location: B11.B2.103
View Authors
- Lauren Pearson (Monash University)
- Nicole Kalms (Monash University)
- Helen Skouteris (Monash University)
- Ben Beck (Monash University)
View Abstract
| Gender inequities in cycling participation persist across low-cycling cities, driven by interacting behavioural, social and environmental barriers. Young adulthood represents a critical window for travel behaviour change, yet few interventions have been systematically designed to address the specific needs of this population. This study applied Intervention Mapping to develop behavioural interventions supporting cycling uptake among young women aged 18–29 years in Melbourne, Australia. Through a mixed-methods design involving a population-representative survey (n=518), in-depth interviews (n=25), and co-design workshops (n=17), we identified four interacting barrier clusters: lack of confidence, lack of skills and knowledge, lack of resources, and concerns about social norms. Four multi-level interventions were co-designed to address these barriers: 1,2,3 Ride (a workshop series with financial incentive), Wheel Wise (gender-sensitive bike shop accreditation), Cycle Social (casual social rides), and Cycle Sundays (regular street closures). Findings demonstrate that young women's cycling barriers operate as an interconnected system requiring complementary interventions across individual, interpersonal, institutional and community levels. This work provides a methodological case for systematically designing cycling interventions within specific population groups experiencing windows of travel behaviour change. |
Paper 152
Understanding pathways to walking in later life: A structural equation modelling approach
Room Location: B11.B2.103
View Authors
- Stephen Greaves (The University of Sydney)
- Ashikur Rahman (The University of Sydney)
View Abstract
| Walking plays a critical role in maintaining physical and mental capacity in later life, yet walking levels among older adults remain uneven. Understanding why some individuals maintain or increase walking as they age, while others do not, is essential for effective policy and intervention design. Informed by socio-ecological frameworks of health behaviour, this study applies structural equation modelling to examine how functional capabilities, perceived barriers to walking, competing modal alternatives and neighbourhood context shape walking for older adults. Data are drawn from a travel and health survey of 2,018 Greater Sydney residents conducted in late 2025, supplemented with published indices of suburb walkability and social disadvantage. Functional limitations emerged as the strongest determinant of walking duration, reflecting the compounding effects of declining physical capacity and reduced tolerance for terrain and distance. In contrast, perceived pedestrian environment barriers were positively associated with walking duration once functional limitations and car convenience were included, consistent with an exposure effect whereby individuals who walk more are more likely to encounter and report barriers. Age was not directly associated with walking with effects manifesting indirectly through increased functional limitations and reduced reporting of environmental barriers, suggesting behavioural adaptation in later life. Neighbourhood walkability was marginally associated with increased walking duration after controlling for other factors. These findings indicate that while walkability remains a critical community-level goal, policies that support functional capacity, confidence and adaptive behaviour are integral to maintaining and increasing walking participation as populations age. |
Paper 3
Regulation in general equilibrium: A quantitative spatial transport model for fare pricing
Room Location: B11.B2.104
View Authors
- Isaac Mann (The University of Sydney)
- David Levinson (The University of Sydney)
View Abstract
| Standard transport appraisal treats fare regulation in isolation, ignoring how price signals reshape urban geography. We bridge regulatory practice and quantitative spatial economics by constructing a Quantitative Spatial Transport Model (QSTM) for Greater Sydney, endogenising household and firm locations in response to transport shocks. Our framework extends the canonical QSM by endogenising commuter mode choice and internalising network externalities through road congestion and the Mohring effect. We apply the model to the 2016 weekly public transport fare cap of $60, evaluating counterfactuals in which the cap is progressively raised. Total social surplus is maximised by removing the cap: current binding caps impose a deadweight loss by subsidising long-distance commuting beyond the point where the marginal social benefits of mode shift outweigh the loss of farebox revenue and the associated marginal excess burden of taxation. Weekly fare caps therefore function as an implicit subsidy for long-distance commuters, and their removal triggers a distinct spatial reorganisation: residents centralise to minimise commuting costs while firms decentralise toward suburban sub-centres to access labour pools priced out of long-distance commuting. Efficient distance-based pricing promotes a compact residential form, encourages polycentric employment, and reduces operator resource costs. |
Paper 22
Risk perception across attributes in cooperative merging: A prospect-theoretic experiment
Room Location: B11.B2.104
View Authors
- Gayasha Samarakoon (The University of Queensland)
- Zuduo Zheng (The University of Queensland)
- Mehmet Yildirimoglu (The University of Queensland)
View Abstract
| This paper examines whether monetary valuation recovers the same cross‑attribute trade‑offs as perceptual equivalence judgments in multi‑attribute decisions involving safety, travel time, and fuel consumption. Using stated‑preference data from a driving‑related experiment, the analysis contrasts certainty‑equivalence (CE) tasks with a perceptual normalisation task in which monetary variation is held constant. Cumulative prospect theory (CPT) is employed as a diagnostic framework to make explicit how heterogeneous attributes are scaled under each elicitation mechanism. A common behavioural framework is used to make cross‑attribute scaling explicit under each elicitation mechanism. The results reveal a clear divergence. CE tasks produce substantial and precisely estimated trade‑offs across attributes, with travel time and fuel retaining a large share of the safety equivalent. In contrast, perceptual normalisation yields strong dominance of safety, with translation parameters for time and fuel collapsing toward behavioural irrelevance. Respondent‑level bootstrap analysis confirms that this collapse is robust to sampling variation. Diagnostics based on stated Most Important Attribute (MIA) indicate limited alignment between declarative importance and perceptual weighting, while safety dominance persists across all groups. These findings suggest that monetary valuation does not merely reveal underlying perceptual trade‑offs but actively constructs commensurability across heterogeneous attributes. Perceptual equivalence, stated importance, and monetised valuation therefore represent distinct behavioural constructs rather than interchangeable measures. The results highlight the importance of accounting for elicitation mechanisms when interpreting cross‑attribute trade‑offs in applied transport analysis. |
Paper 27
Toward carbon neutrality: Feasibility of an ETS for Sri Lanka’s transport sector
Room Location: B11.B2.104
View Authors
- Sajini Nawela (University of Moratuwa)
- Loshaka Perera (North Dakota State University)
View Abstract
| The global transport sector is a significant contributor to greenhouse gas (GHG) emissions and a key driver of climate change, with road transport responsible for nearly 70% of transport-related energy emissions worldwide. In Sri Lanka, transportation accounts for around 15% of total national GHG emissions, highlighting the sector as a priority area for achieving the country’s carbon neutrality target by 2050. Emission Trading Schemes (ETS) have gained international recognition as an effective market-based mechanism for reducing emissions by creating financial incentives for cleaner practices, while also encouraging technological advancement and cost efficiency. Despite successful implementation in several sectors globally, the application of ETS to Sri Lanka’s transport sector has received limited attention. This study examines the feasibility of introducing an ETS framework for Sri Lanka’s land transportation sector by analyzing international models and assessing their relevance to local conditions. A mixed-methods methodology was adopted, combining qualitative data from expert interviews with quantitative findings from public surveys. Results indicate strong support for the adoption of ETS, particularly through realistic emission caps for high-emission vehicle groups such as buses and taxis, and through alignment with proven international practices. Survey participants also emphasized the importance of fairness in implementation and the allocation of ETS revenues toward improving public transport systems. The study offers practical policy recommendations for designing an ETS suited to Sri Lanka’s transport sector, supporting national sustainability goals and the transition to a low-carbon economy. |
Paper 67
Evaluating transport options across uncertain futures: Integrating stratified multi-criteria analysis into ATAP and Infrastructure Australia guidance
Room Location: B11.B2.104
View Authors
- Miguel Loyola (The University of Western Australia)
- Tony Arnold (The University of Sydney)
View Abstract
| Transport appraisal in Australia relies on multi-criteria analysis (MCA) at two critical stages: early options filtering within the ATAP Framework and project prioritisation for the Infrastructure Australia Infrastructure Priority List. Both frameworks recognise the limitations of static criteria weighting, with ATAP cautioning against the single weighted score and Infrastructure Australia explicitly requiring scenario sensitivity testing. However, neither framework provides a structured method for varying criteria weights across plausible futures before a ranking is produced. This paper introduces Stratified Multi-Criteria Analysis (SMCA) as a practical enhancement to both frameworks. SMCA extends conventional MCA by assigning criteria weights separately across structured future scenarios, each representing conditions that could plausibly prevail over the life of a transport asset. Criteria weights are elicited separately under each scenario, alternative performance scores are evaluated within each scenario, and probability-weighted expected scores and rankings are derived. The approach is illustrated using a Perth metropolitan transport policy context, evaluating four competing policy options across three uncertain events: an election changing environmental policy settings, an economic recession, and accelerated population growth. The paper shows that the conventional MCA ranking can differ from the probability-weighted ranking in ways that are directly relevant to long-term investment decisions, and that SMCA provides the robustness diagnostics that Infrastructure Australia already requires but current guidance does not operationalise in a scenario-dependent weighting procedure. The method is positioned as a guideline-compatible enhancement to existing Australian appraisal practice, not a replacement for cost-benefit analysis. |
Paper 18
Missing the innovation passenger: Methodological blind spots in high-speed rail research
Room Location: B11.B3.101
View Authors
- Nicholas Hurley (Monash University)
- Liton Kamruzzaman (Monash University)
- Graham Currie (Monash University)
View Abstract
| Measuring the travel behaviour of high-income knowledge workers remains a persistent challenge across the transport sector; these time-poor individuals possess a high opportunity cost of time, making them highly resistant to participation in traditional travel surveys. For High-Speed Rail (HSR), this sampling bias creates a critical disconnect between policy intent and empirical reality. Whilst the economic justification for HSR increasingly relies on the theory of agglomeration and the facilitation of the knowledge economy, the methodological toolkit used to evaluate these benefits remains largely rooted in traditional transport utility metrics that fail to capture this pivotal demographic. This paper synthesises 62 empirical studies to examine whether existing data collection approaches are equipped to capture the relational mechanisms central to the knowledge economy. The analysis reveals a systemic sampling paradox: the specific economic agents driving regional innovation are systematically under-represented by data collection designs that prioritise high-burden, retrospective intercept surveys. Consequently, the social mechanisms of innovation, such as weak tie formation and serendipity, remain unobserved. Synthesising these findings against emerging capabilities in digital mobility sensing, the paper proposes a Context-Aware Hybrid Framework. This methodological approach moves beyond the station-to-station paradigm, offering a blueprint for combining low-barrier recruitment with time-triggered digital engagement. By minimising respondent burden whilst enhancing temporal and social granularity, this framework provides a pathway to capture the elusive innovation worker, thereby allowing for the rigorous empirical validation of the HSR agglomeration hypothesis. |
Paper 66
Stepping up to stepping out: Factors associated with independent mobility licences in Australia
Room Location: B11.B3.101
View Authors
- Alexa Delbosc (Monash University)
- Maryam Jafari (Monash University)
- Emily Moylan (The University of Sydney)
- Jennifer Kent (The University of Sydney)
View Abstract
| Children’s independent mobility is constrained by the permissions granted or denied by their parent or caregiver. For this reason, many studies of children’s independent mobility focus on understanding when parents grant an Independent Mobility Licence (IML). These licences can include permission to cross main roads unsupervised, travel home from school without an adult or go out after dark. Many of those studies sum or otherwise combine different licences into a combined indicator of a child’s potential to travel independently. The aim of this paper is to test which characteristics of a child, their family or their environment are associated with being granted a specific IML, rather than assume that these factors apply equally across all licence types. We further explore whether these patterns differ between children in primary vs secondary school. We explore this aim using the Australian School Travel Survey, a nation-wide survey capturing the travel and mobility licence of 7,909 school-age children in Australia. A set of binary logistic regression models predict whether or not a child is granted licence to cross main roads, travel home from school, travel to destinations other than school, use public transport or go out after dark unsupervised by adults. We find that a child’s age and gender are the only variables that are consistent across all models, with older children and boys more likely to be granted a given licence than younger children or girls. Beyond that, the variables predicting individual licences varied depending on the licence and school level. The findings suggest that mobility licences should not be summed together into a single index. The models also reveal patterns that suggest future researchers should look in more depth at the socio-cultural context of parenting and how it shapes the affordances that parents grant their children to travel independently. |
Paper 98
A review of stated preference findings on travel attributes for High-speed rail
Room Location: B11.B3.101
View Authors
- Sathvik Gadiraju (Monash University)
- L Hai Vu (Monash University)
- Liton Kamruzzaman (Monash University)
View Abstract
| High-speed rail (HSR) is increasingly proposed as a strategic intercity transport investment, but its substantial capital cost makes reliable demand forecasting essential. In the absence of historical data for non-existing alternatives, stated preference (SP) and stated choice (SC) experiments have become the dominant tool for estimating HSR mode choice and demand. This paper presents a systematic review of 32 SP/SC studies, published between 1992 and 2026, that estimate intercity mode choice or passenger demand with HSR. Studies were identified through Scopus and Web of Science. The review synthesises evidence on the attributes shaping HSR mode choice, including travel time, travel cost, frequency, reliability, and comfort, and highlights how findings vary across contexts. Gaps are identified, including the redundancy between frequency and waiting time specifications, the limited treatment of onboard amenities, and the absence of environmental attributes despite HSR's emissions advantage. These findings offer guidance for future SP designs in HSR demand modelling. |
Paper 115
NSW statewide active transport survey – What behaviour matters where?
Room Location: B11.B3.101
View Authors
- Sara Haider (The University of Sydney)
- John Nelson (The University of Sydney)
View Abstract
| Active transport is widely recognised for its environmental, health, and economic benefits, yet walking and cycling uptake remains low in Australia. A challenge in planning is the assumption that active transport interventions provide similar benefits across contexts, despite substantial differences in population, settlement, and environmental conditions. This paper examines the behavioural factors influencing walking and cycling uptake across New South Wales (NSW), and whether those factors vary across different areas in the state. The analysis draws on a statewide survey of 2,293 respondents across metropolitan and regional NSW, capturing demographic, contextual, behavioural, and infrastructure-related factors. Hierarchical regression is used to examine the relationship between these behavioural factors and reported frequency of walking and cycling, progressing from opportunity factors to demographic, attitudinal, and reported use factors. The findings were triangulated using gradient boosting and SHapley Additive exPlanations (SHAP). The results show that cycling is more strongly shaped by access-related, attitudinal, and infrastructure comfort factors, with greater variability across regional and lower-density contexts. Walking is more strongly driven by basic feasibility and motivational factors, with less variation across the state. These findings show that active transport behaviour is not shaped uniformly across contexts, and that cycling requires more context-sensitive intervention. The paper contributes evidence to support more targeted and locally responsive active transport planning and policy. |
Paper 9
Exploring workforce transitions in brownfield automated metro development: A practice and literature review
Room Location: B11.B3.103
View Authors
- Vikrant Sharma (Monash University)
- Graham Currie (Monash University)
- Greg Bamber (Monash University)
View Abstract
| Automation of metro systems has been growing significantly globally. This includes greenfield projects (newly built networks) and brownfield projects (conversion from an old network). Focussing on the latter, this paper aims at exploring the impact of automation on workforce employment and workforce transition practices through a practice and literature review. Current trends of automation are also reviewed including brownfield projects currently in development. Every brownfield project grapples with the question – where will the existing staff go? To assess this, the four completed brownfield projects are reviewed including Nuremberg, Paris Metro Lines 1 and 4 and Lyon Metro Line B. Impacts of automation on workforce employment and workforce transition strategies are synthesised. The research found that between 2000 and 2023, the number of automated metro systems increased from 18 to 115. Interestingly only 4 involved brownfield systems and each of these were in countries with strong workforce protection policies (Germany and France). Of the four brownfield metro systems explored; workforce employment in Nuremberg declined from 192 to 179 (-13) and from 219 to 40 (-179) in Paris Metro Line 1. Data for other brownfield sites are unavailable. Workforce transitions in all the cases explored involve a no redundancy policy in both cases. Displaced workers were employed in other rail corridors or retrained. In the Paris Metro Line 1 case, compensation for dislocation associated with automation was paid. The paper explores the implications for future research and practice. |
Paper 44
A Node-Place-Network Model for evaluating TOD and hub stations using passenger flow-based centrality
Room Location: B11.B3.103
View Authors
- Prima Romadhona (Universitas Islam Indonesia)
- Ronghui Liu (University of Leeds)
- Chandra Balijepalli (University of Leeds)
View Abstract
| This study develops and applies a Node–Place–Network Model (NPNtM) to evaluate Transit-Oriented Development (TOD) and hub roles within a metropolitan commuter rail system. Building on the traditional Node–Place Model (NPM), the research incorporates a third “network” dimension derived from weighted centrality measures using passenger flow data from Automated Fare Collection (AFC) systems. The methodology models the rail network in P-space and integrates network centralities to capture stations’ functional roles within the wider system. The framework distinguishes between TOD-hub, TOD non-hub, and non-TOD, enabling more nuanced assessment of performance. The results reveal that hub status alone does not determine performance. Instead, centrality is jointly shaped by demand intensity, spatial position, and service provision. While a small number of TOD-hubs dominate network flows, most stations exhibit relatively similar accessibility levels but limited bridging roles. The NPNtM classification identifies four station types, ranging from high-performing core hubs to peripheral low-performing TOD, demonstrating that stations with similar local characteristics can differ significantly when network embedding is considered. The study shows that integrating passenger-weighted network analysis into TOD evaluation provides a more comprehensive understanding of station performance, enabling better identification of strategic TOD and more targeted policy interventions in developing metropolitan contexts. |
Paper 46
Detection of accessibility features for people with disabilities at bus stops: A comparison of YOLOv8 and Faster R-CNN models on Google Street View imagery
Room Location: B11.B3.103
View Authors
- Tuo Mao (University of Technology Sydney)
- Hin Tai Chan (University of Technology Sydney)
View Abstract
| Public transport accessibility is essential for inclusive and equitable urban planning, particularly for people with disabilities who depend on bus services. Traditional bus stop accessibility audits rely on manual inspections, which limits efficiency at scale. This study proposes an automated framework that uses Google Street View imagery and object detection models (YOLO and Faster R-CNN) to assess accessibility features relevant to people with disabilities. The approach could transform accessibility assessment by supporting evidence-based urban planning, transport policy, and the daily travel needs of bus users. By automatically detecting key features (e.g., seating, shelters, and bus stop signs), the system provides a more comprehensive view of stop-level accessibility and supports the development of a more inclusive public transport network. Despite limited training data, the models achieved promising performance in both object detection and dimensional measurement. |
Paper 119
Assessing Public Transport Accessibility Levels (PTAL) for regional transport planning: The case of Sunshine Coast, Queensland
Room Location: B11.B3.103
View Authors
- Louise Mikkelsen (University of the Sunshine Coast)
- Lynette Cheah (University of the Sunshine Coast)
View Abstract
| This study examines the applicability of the Public Transport Accessibility Level (PTAL) metric for transport and land-use planning in rural and regional contexts. The suitability of PTAL for regional transport planning is analysed for the case study of the Sunshine Coast, Queensland. The PTAL methodology was applied across 119 suburbs using public transport data and walking time metrics. Results indicate that public transport accessibility across the region is uniformly low, with the majority of suburbs classified within the lowest PTAL bands. Whilst PTAL effectively highlights broad spatial inequities and communicates the scale of transport disadvantage, its effectiveness is limited when applied in dispersed, regional environments, where accessibility is fundamentally different from urban environments. To address this, a locally calibrated accessibility classification was developed alongside the standard PTAL bands, enabling more meaningful comparison between suburbs while preserving the original methodology. The findings suggest that PTAL is more valuable as a strategic benchmarking and screening tool in regional contexts. |
Paper 12
Relative travel time and public transport competitiveness for NSW workers: A model paper
Room Location: B11.B3.104
View Authors
- Gazal Kapoor (Transport for NSW)
- Hasini Abeywickrama (Transport for NSW)
- Lluvia Ding (Transport for NSW)
- Luke Freeman (Transport for NSW)
- James Hilder (Transport for NSW)
- Dat Nguyen (Transport for NSW)
- Srikanta Patra (Transport for NSW)
View Abstract
| Understanding the competitiveness of public transport (PT) relative to private vehicles is essential for developing effective, evidence-based transport policy in New South Wales (NSW). This study examines revealed mode choice for NSW commuters, integrating Census Journey to Work (JTW) data with public transport timetable information and private vehicle travel times derived from HERE travel time data. The findings show that PT market share is highly sensitive to the PT competitiveness calculated as the ratio of Public transport Journey Time over Private Vehicle journey Time. |
Paper 20
Beyond physical mobility: A multi-layer urban simulation model incorporating digital accessibility and social dynamics
Room Location: B11.B3.104
View Authors
- Mustafa Mutahar (Toyohashi University of Technology)
- Nao Sugiki (Toyohashi University of Technology)
- Fumitaka Kurauchi (Gifu University)
- Kojiro Matsuo (Toyohashi University of Technology)
View Abstract
| Cities are changing. The way people access services, interact with each other, and decide where to live is no longer shaped by physical infrastructure alone; digital connectivity and social relationships are also playing an increasingly important role. Yet most urban simulation models still treat the city as if digital accessibility does not exist, focusing almost entirely on roads, transit, and land use. This leaves a significant blind spot in how we understand and plan for long-term urban change. This study addresses that gap by developing a multi-layer Social Dynamics Simulation (SDS) model that brings together three dimensions of urban life: the physical network of transportation and facilities, the virtual network of digital services, and the social network of interpersonal relationships. Rather than treating these as separate systems, the model integrates them into a single framework through an integrated accessibility index, a measure that reflects both where people can physically go and what they can access digitally. The index is grounded in a service choice model estimated from a large-scale survey of 6,210 respondents, capturing how individual characteristics such as age, employment status, and the importance placed on workplace and online relationships influence whether people choose physical or digital services. Applying the model to a virtual city over a 30-year simulation period reveals several important findings. Digital accessibility does not replace the pull of good transport connections; the population still concentrates around transit nodes, but it does improve the traffic conditions. Households with stronger digital service tendencies are more willing to live in areas with lower physical accessibility, producing a more spatially dispersed city than conventional models would predict. At the same time, digital service substitution helps moderate traffic demand, while public transport adapts by increasing service frequency along high-demand corridors. Accessibility gaps between population groups also emerge clearly: elderly residents without cars tend to cluster in highly accessible areas, while those with vehicles enjoy greater residential location choices. These findings from Japan suggest that as cities navigate rapid digital transformation, planning frameworks that account only for physical mobility are missing a crucial part of urban space. The proposed model offers a practical tool for evaluating policies at the intersection of transport, digital infrastructure, and social equity. |
Paper 95
An agent-based land use and transport interaction framework for modelling the temporal and spatial effects of high-speed rail on city population dynamics
Room Location: B11.B3.104
View Authors
- Saeid Adli (Monash University)
- Liton Kamruzzaman (Monash University)
- Graham Currie (Monash University)
- Le Hai Vu (Monash University)
- Dong Xiao (Monash University)
View Abstract
| This paper presents an agent-based Land Use and Transport Interaction (LUTI) framework for evaluating the temporal and spatial effects of high-speed rail (HSR) on city population dynamics in Australia. We argue that the proposed MATSim and UrbanSim coupled framework is uniquely capable of capturing the ambiguous, spatially differentiated, and temporally unfolding population effects of HSR. These are effects that traditional aggregate models systematically miss. The framework integrates three core components: (1) an agent-based transport model (MATSim) that simulates individual travel behaviour, mode choice, and network dynamics; (2) an agent-based land use model (UrbanSim) with disaggregated sub-models for household location choice, employment location choice, real estate pricing, and real estate development; and (3) an iterative coupling protocol that captures the dynamic feedback between transport accessibility and urban development over a 10 to 30 year horizon. We formalise the causal chain from HSR investment through generalised cost reduction, accessibility improvement, productivity gains, and household migration to a new spatial equilibrium. The temporal dimension is central to the framework's design, distinguishing short-run travel responses (0 to 2 years), medium-run firm relocation (2 to 10 years), and long-run population redistribution (10 or more years). We demonstrate how Australian data sources (ABS Census, BITRE intercity flows, SA4-level GDP, cadastral records) support model calibration, and justify why agent-based approaches for both transport and land use are superior to traditional four-step and aggregate models for capturing HSR-induced city size dynamics. |
Paper 107
The governance tale of two cities: A comparative analysis of land use and transport integration in Australia
Room Location: B11.B3.104
View Authors
- Rafsana Hossain (Monash University)
- Liton Kamruzzaman (Monash University)
- Elizabeth Taylor (Monash University)
- Tooran Alizadeh (The University of Sydney)
View Abstract
| Land use and transport integration initiatives often produce contrasting outcomes across cities, even within broadly similar governance systems. This paper examines why such differences emerge through a comparative analysis of two Australian cases: Newcastle, New South Wales, a post-industrial restructuring city, and Tarneit, Victoria, a rapidly growing outer-metropolitan growth area. Using a governance framework as an analytical lens, the study investigates how governance attributes were enacted, prioritised, omitted, and negotiated across different institutional and urban contexts. The analysis draws on document review and stakeholder interviews to examine governance processes across project phases in both cases. The findings reveal substantial divergence in the operationalisation of governance attributes. In Newcastle, strong state leadership, institutional coordination, and alignment across actors supported relatively coherent implementation despite community pressures. In Tarneit, tensions between state and local priorities, limited local government capacity, weakened community influence, and strong developer interests constrained integrated governance responses. The paper argues that governance outcomes are shaped not simply by the presence of governance frameworks, but by how governance attributes are activated across phases, interact with institutional barriers, and align with local contexts and implementation conditions. The findings highlight the importance of context-sensitive approaches to land use and transport integration in Australian cities |
12:40 – 1:20
Lunch
1:30 – 3:00
Paper presentation period 6
Paper 113
A systematic review of surrogate safety measures in work zone safety analysis
Room Location: B11.B2.101
View Authors
- Tharindu Ranawaka (Deakin University)
- Davies Huw (Coventry University)
- Christophe Bastien (Coventry University)
- Ashim Debnath (Deakin University)
View Abstract
| Ensuring safe and efficient transportation within work zones is crucial for road network authorities. While previous research has focused on crash analysis based on collision data, work zone-specific studies require customised approaches to analyse crash risks. This systematic review examines the use of surrogate safety measures in work zone safety assessments. The review identified 41 relevant studies addressing crash risk assessments using surrogate safety measures in work zones. The review findings showed that time-to-collision remains the dominant surrogate safety measure in work zone safety studies, particularly for rear-end and merging conflict analyses. It further finds that combining time‑based measures with kinematic and encroachment‑based indicators is a common approach to capture the diverse conflict mechanisms observed in work‑zone environments. To overcome these limitations, innovative measures such as the Lane Change Risk Index and Time to Collision in the Work Zone have been introduced. Furthermore, considering factors beyond vehicle interactions, such as traffic volume, proportion of large vehicles, work zone lengths, vehicle mass, lane-changing behaviour, and geometric design, provides a more comprehensive evaluation of work zone safety. The findings emphasise the need for customised interventions and periodic systematic reviews to incorporate new developments and address research gaps in work zone crash analyses. The findings of this systematic review support the tailored application of surrogate safety measures in work zones, enabling more accurate crash risk assessment and the development of targeted safety strategies. |
Paper 120
Smart road accident detection from surveillance videos using transformers
Room Location: B11.B2.101
View Authors
- Tanu Singh (Indian Institute of Technology Kanpur; La Trobe University)
- Pranamesh Chakraborty (Indian Institute of Technology Kanpur)
- Long Truong (La Trobe University)
View Abstract
| Traffic accidents are a leading global cause of death, increasing with population growth, urbanization, and motorization, raising concerns about traffic surveillance effectiveness. Traditional computer vision methods for accident detection struggle with limited spatiotemporal understanding and poor cross-domain generalization. Few studies have employed recurrent architectures, including Long Short-Term Memory (LSTM) networks and Convolutional Neural Network-Recurrent Neural Network (CNN-RNN) hybrids, to model temporal information. However, these approaches often struggle to capture long-range temporal relationships and are computationally intensive. Recent developments in transformer architectures have demonstrated superior performance in modeling global spatial and temporal dependencies while enabling parallel computation. Nevertheless, using transformers for encoding both spatial and temporal feature increases computational demands, which limits real-time application. To address these challenges, this study adopts a hybrid approach that balances efficiency and performance. However, applying these approaches to traffic accident detection is often limited by existing small and non-diverse datasets, hindering the development of robust and generalizable system. To address this gap, we curated a comprehensive and balanced dataset that captures wide spectrum of traffic environments, accident types, and contextual variations. Utilizing the curated dataset, we propose an accident detection model based on a transformer architecture using pre-extracted spatial video features. The architecture employs convolutional layers to extract local correlations across diverse patterns within a frame, while leveraging transformers to capture sequential-temporal dependencies among the retrieved features. Moreover, most existing studies neglect the integration of motion cues, which are essential for understanding dynamic scenes, especially during accidents. In this study, multiple methods for incorporating motion cues were evaluated to identify the most effective strategy. Among the tested incorporating methods, concatenating RGB features with optical flow achieved the highest accuracy at 88.3\%. The results were further compared with vision language models to assess the effectiveness of the proposed method. |
Paper 122
An extreme value model for estimating vulnerable road users expected crashes from traffic conflicts on slip lanes
Room Location: B11.B2.101
View Authors
- Harpreet Singh (Queensland University of Technology)
- Shimul Md Mazharul Haque (Queensland University of Technology)
- Andrew Burbridge (Department of Transport and Main Roads)
View Abstract
| Traffic conflict modelling using extreme value models has emerged as a key approach for estimating expected crashes from observed traffic conflicts. However, proximity-based conflict definitions alone may lead to biased estimates because they can include controlled interactions that are spatially or temporally close but do not necessarily indicate elevated crash risk. In contrast, evasive action-based conflicts provide a more behaviourally meaningful alternative by focusing on interactions in which road users respond to perceived risk, such as through vehicle deceleration or speed adjustment. Despite recent methodological advances, limited research has incorporated evasive action-based conflicts for crash risk estimation, particularly for vehicle–vulnerable road user interactions. To address this gap, the present study integrates pre-event evasive action indicators into the identification of safety critical conflicts and applies a Peak over threshold extreme value model to estimate expected crash frequency. Video data were collected from 30 sites across Brisbane, Townsville, and Cairns, Australia, covering four slip lane configurations: unmarked slip lanes, zebra controlled slip lanes, raised pedestrian crossings, two-aspect signalised slip lanes, and exclusive left-turn lanes, with six sites per configuration. Road user trajectories were extracted using Artificial Intelligence-based video analytics, and model performance was evaluated by comparing predicted and observed conflict frequencies. The findings indicate that evasive action-based conflicts can reasonably capture the occurrence of extreme conflict events across diverse slip-lane configurations. Annual expected crash frequency estimates revealed significant differences in safety across slip lane configurations. Raised pedestrian crossings had the lowest estimated expected crash frequency (0.078), followed by zebra-controlled slip lanes (0.604), while unmarked slip lanes showed the highest risk (0.811). Among signal-controlled facilities, two-aspect signalised slip lanes had a higher expected crash frequency (0.307) than exclusive left-turn lanes (0.255). Overall, the findings demonstrate the importance of incorporating evasive action indicators into extreme value theory-based conflict modelling and provide evidence to support proactive safety assessment and safer infrastructure design. |
Paper 123
Travel time impacts of safe speed limits: Evidence from Merri-bek, Australia
Room Location: B11.B2.101
View Authors
- Vidumini Daluwathu (The University of Sydney)
- Mulle Gamage (The University of Sydney)
- Emily Moylan (The University of Sydney)
- John D. Nelson (The University of Sydney)
View Abstract
| Safe speed limit policies are increasingly implemented on local residential roads to improve road safety, yet concerns remain regarding their impacts on travel times and everyday mobility. Empirical evidence on network‑wide travel time effects of reduced speed limits in urban environments is limited. This paper evaluates the impacts of a large‑scale speed limit reform implemented in Merri‑bek, Australia, on travel time. Using high‑resolution connected vehicle data, a before‑and‑after analysis is conducted to examine changes in journey travel times following the reduction of local road speed limits from 50 kph to 40 kph. Trips are classified by spatial relationship, road hierarchy, and time‑of‑day, and impacts are assessed using unit travel time and a percentile‑based framework to capture distributional effects across different trip types. Results indicate that overall travel time impacts are modest. During weekday peak periods, congestion dominates operating conditions, resulting in minimal changes to typical journey times. Larger, but still moderate, increases are observed during off-peak periods, when vehicles operate closer to posted speed limits. Across both periods, impacts are concentrated among faster vehicles, while median travel times remain largely stable. Local roads, which often have short block lengths, driveways, on-street parking and crossings, experience relatively limited changes compared with other parts of the network. These findings suggest that safe speed limit policies can mitigate the most dangerous speeds without imposing substantial travel time penalties on most trips, thus informing policy debates on safety and urban mobility. |
Paper 125
FASTrack: A co-designed planning support tool for cycling infrastructure scenario assessment in regional areas
Room Location: B11.B2.103
View Authors
- Ryan Turner (RMIT University)
- Afshin Jafari (RMIT University)
- Sapan Tiwari (RMIT University)
View Abstract
| Regional cities are increasingly expected to deliver safer and more connected active transport networks, yet many lack the data, modelling capabilities and decision-support tools needed to assess cycling infrastructure options in a consistent way. This paper presents FASTrack, a web-based planning support tool developed to help practitioners create, test and interpret cycling infrastructure scenarios. Using Greater Bendigo as a pilot case study, FASTrack connects an attribute-rich transport network, a regional activity-based demand model, cycling route choice analysis and an active transport cost benefit analysis framework. The tool allows users to draw or edit proposed cycling infrastructure through a browser-based ‘Scenario Maker’ and then interpret scenario outputs through a dashboard-based ‘Scenario Viewer’. Outputs include changes in network conditions, cycling exposure, mode shift, emissions and economic appraisal metrics. This paper describes the rationale for FASTrack, the co-design process used to shape its functions, the modelling workflow behind the tool, and the current scope and limitations of the prototype. The contribution of FASTrack is its translation of transport modelling and appraisal methods into a practical planning environment that can support early-stage option testing, stakeholder discussion and investment prioritisation. The paper argues that co-designed, scenario-based tools can help bridge the gap between active transport modelling research and the practical needs of local and state government planning, particularly in regional contexts where data and modelling resources are often limited. |
Paper 146
From consideration to action: Identifying amenable population subgroups for sustainable mode shift during the 2026 Australian fuel price crisis
Room Location: B11.B2.103
View Authors
- Lauren Pearson (Monash University)
- Ana Luiza de Sa (Monash University)
- Robyn Gerhard (Monash University)
- Jamie Abrahams (Monash University)
- Hossein Nosratzadeh (Monash University)
- Toby Cumming (Monash University)
- Ben Beck (Monash University)
View Abstract
| In early 2026, disruption to global oil supply resulted in sharp fuel price increases in Australia. Acute price shocks may create windows of opportunity for sustainable travel behaviour change, yet which populations are positioned to adapt during crises remains poorly characterised. A population-representative cross-sectional online survey of Australian adults conducted in April 2026 (n=2,177) captured sociodemographics, pre-crisis travel behaviour, considered and enacted mode shifts, and supports for sustained change. Three-quarters of respondents (74.6%) had considered or begun changing how they travel. Walking was the most commonly adopted mode (20.4%) and trip reduction the most common adaptation (40.7%). Behaviour change was substantially more common among those experiencing financial hardship (84.6% vs 55.6%; p<0.001) and younger respondents. Free public transport induced new or increased use among 44.7% in Victoria but only 24.6% in Tasmania. Just 7.9% indicated they would definitely continue new behaviours if prices fell, while 37.6% were unsure. Using explainable machine learning, younger age and financial hardship emerged as dominant predictors of amenability to mode shift. Supervised clustering of SHAP values identified three distinct archetypes: financially pressured young adults, employed urban professionals, and regional residents—with a subset of regional respondents amenable to public transport but lacking service access. The fuel crisis has produced substantial gaps between considered and enacted behaviour change, with amenability concentrated in identifiable subgroups whose drivers differ by mode. |
Paper 147
Public orientations toward distributive principles for walking equity: A latent class analysis
Room Location: B11.B2.103
View Authors
- Ashikur Rahman (The University of Sydney)
- Stephen Greaves (The University of Sydney)
- Chinh Ho (The University of Sydney)
View Abstract
| Walking equity, as a fair resource allocation objective, is increasingly recognised as a multi-disciplinary concern in transport planning, urban planning, public health, and climate-responsive urban design. Nonetheless, the definition of equitable walking resource distribution remains a matter of argument among researchers. The same pedestrian intervention can be justified or critiqued through a separate lens of distributive justice: maximising overall welfare, reducing inequalities between people and places, or upholding a minimum threshold of walkability for all. However, these normatively distinct distribution principles are often diluted into a single equity language in active transport equity research. This paper develops a survey-based framework to explore people’s latent walking equity orientations in Greater Sydney across six dimensions or perspectives: accessibility, built environment, health, safety, affordability, and environmental exposure. Drawing on distributive justice theories, this paper operationalised utilitarianism, egalitarianism, and sufficientarianism as the distribution principles through six walking-equity choice tasks. The unobserved walking equity orientations were identified from respondents’ similar response patterns using latent class analysis. A five-class solution emerged as the most defensible model as a balance between statistical fit and class interpretability. People’s equity standpoints were observed to vary across different walking-equity perspectives. Conditional response probabilities were used to interpret and semantically label the latent equity classes as Non-Utilitarian Mix (the largest cluster), Non-Sufficientarian Mix, Strong Egalitarian, Strong Sufficientarian, and Strong Utilitarian (the smallest cluster). The paper applied a bias-adjusted three-step model to explore the associations between six socio-demographic covariates and the five equity classes. At the variable-block level, age, gender, ethnicity, and residence duration were statistically significant with class membership, while income was borderline significant, and education was not significant. However, a closer investigation of the effect-coded coefficients indicated a modest to low association between most of the socio-demographic categories and the equity classes, suggesting a more intuitive interpretation of public equity orientations as a latent normative attribute, rather than deducible by socio-demographics. |
Paper 155
More than infrastructure: How skills and social support enable e-bikes
Room Location: B11.B2.103
View Authors
- Jennifer Kent (The University of Sydney)
- Lara Mottee (Queensland University of Technology)
- Melanie Crane (The University of Sydney)
- Nicole Peel (Western Sydney University)
View Abstract
| This paper presents qualitative findings from an evaluation of the Transformers e-bike subsidy program, a community-delivered initiative targeting low-income residents of Western Sydney. Drawing on in-depth interviews with 23 participants conducted during an eight-week subsidised e-bike trial, the study examines how access to e-bikes—when combined with structured skills training, guided group rides and peer support—shaped mobility practices, confidence, wellbeing and everyday travel behaviour in a car-oriented suburban context. Participants described substantial benefits, including reduced reliance on private cars and public transport, increased uptake of active transport for everyday trips, and marked gains in confidence, skills and road awareness. These changes were frequently experienced as transformative, particularly among participants with low initial confidence, older adults and women. Importantly, the analysis demonstrates that the financial subsidy alone was insufficient to generate these outcomes. Instead, behaviour change was enabled through the integration of training, social support and opportunities to experiment with cycling in real-world conditions. The paper also identifies persistent structural, social and financial barriers that constrained the full integration of e-bikes into daily life, including unsafe or incomplete cycling infrastructure, lack of secure bicycle parking, difficulties integrating e-bikes with public transport, and the affordability of e-bike ownership after the trial. By foregrounding lived experience and mechanisms of change, this paper contributes qualitative evidence to debates on transport equity and active transport policy, demonstrating the conditions under which e-bike subsidy schemes can support sustained behavioural change in low-density, car-dependent cities. |
Paper 106
Beyond the four-step model: Structural differences and a data integration framework for activity-based transport
Room Location: B11.B2.104
View Authors
- Supun Perera (The University of Sydney)
- Michael Bell (The University of Sydney)
View Abstract
| Transport modelling is transitioning from traditional four-step approaches toward Activity-Based Models (AcBMs), reflecting a broader shift from aggregate flow-based representations of travel demand toward behaviourally grounded activity participation frameworks. Unlike conventional models, AcBMs represent travel as the outcome of interconnected decisions relating to activity participation, scheduling, destination selection, and mode choice, constrained by time, accessibility, and household resources. This transition is being accelerated by advances in computational modelling, machine learning, and the increasing availability of large-scale mobility datasets. However, practical implementation remains constrained by fragmented and poorly integrated data ecosystems. Existing datasets provide only partial representations of travel behaviour, with behavioural surveys offering depth but limited scale, population datasets providing coverage but limited behavioural detail, and passive mobility datasets capturing large-scale movement patterns but lacking contextual interpretation. This paper examines the structural differences between conventional four-step models and AcBMs and demonstrates how these differences fundamentally reshape data requirements. A structured data integration framework is proposed that aligns complementary datasets with core AcBM components across behavioural, spatial, temporal, and network dimensions. The findings highlight that advancing AcBMs depends not only on methodological innovation, but also on the development of interoperable and integrated data ecosystems capable of supporting behaviourally realistic transport modelling. |
Paper 124
Tour defence: Extensive hypothesis testing in safety-augmented travel demand models
Room Location: B11.B2.104
View Authors
- Zeke Ahern (Queensland University of Technology)
- Amalan Mahendran (Queensland University of Technology)
- Alexander Paz (Queensland University of Technology)
View Abstract
| Developing activity travel schedules requires several interdependent decisions, including activity participation, duration, timing, destination, and mode choice. These decisions are shaped by individual characteristics, spatial context, and an individuals knowledge of the transport environment. Among the many factors influencing travel behaviour, perceived and objective safety plays a critical role. Travelers may avoid certain modes, choose safer routes, or avoid certain areas, even when less safe alternatives offer advantages such as shorter travel times or attractive zonal features. Conventional activity-based models typically encode safety only implicitly, treating the risk of crash as an outcome rather than a behavioural input. This paper introduces a hierarchical activity-based model for South-East Queensland (SEQ) in which expected crash frequency is explicitly incorporated as a utility component across all relevant decision layers. The framework generates full daily activity–travel plans for a synthetic population of 3.7 million residents distributed across 329 zones (Statistical Area Level 2), producing origin–destination flows suitable for dynamic network assignment. Coordinated daily activity pattern classification, activity frequency, day-duration budgeting, per-activity duration, schedule placement, mode and destination choice is estimated using the Queensland Household Travel Survey and applied to the full synthetic population. Safety terms derived from crash-prediction functions are embedded within the mode-choice and destination-choice utilities, allowing the model to capture safety’s influence on behaviour alongside travel time, distance, and cost. A genetic-algorithm-based feature selection procedure supports and extends analyst-driven variable specification by conducting extensive hypothesis testing across all model stages, identifying the person attributes and network characteristics most influential in each decision layer. Controlled experiments comparing the safety-integrated model with a baseline mode, quantifies how crash-risk exposure reshapes mode shares, destination patterns, and overall daily travel behaviour. The results show that omitting safety systematically misestimates active-mode use and introduces spatial bias in destination predictions for vulnerable road users. |
Paper 139
Proxy-based importance sampling for ActivitySim location choice: Accuracy, distance bias and runtime
Room Location: B11.B2.104
View Authors
- Tom Stephen (Outer Loop Consulting)
- Jan Zill (Outer Loop Consulting)
View Abstract
| This paper evaluates random sampling and proxy-based importance sampling for non-mandatory tour destination choice in ActivitySim using both probability-based error metrics and trip-length diagnostics. In the tested regime, distance-based importance sampling substantially reduces dispersion error and RMSE relative to random sampling and reproduces full-set trip-length outcomes much more closely across the sample sizes examined. A runtime comparison also shows that sampling 30 alternatives retains large computational savings relative to the evaluation of the full 1,449-alternative choice set. The results indicate that, in application, the practical value of sampling depends primarily on the sampling method rather than on sample size alone, and that a commonly used sample size of 30 alternatives is defensible in this setting when paired with importance sampling. |
Paper 145
Machine learning-based spatio-temporal demand prediction for taxi and ride-hailing systems: A systematic review of contextual information representation and future directions
Room Location: B11.B2.104
View Authors
- Mohamed Marzook Fathima Safiya (Robert Gordon University)
- Ppg Dinesh Asanka (University of Kelaniya)
View Abstract
| Accurate spatio-temporal demand prediction is fundamental to the efficient operation of taxi and ride-hailing systems, supporting operational decisions such as fleet allocation, dispatch planning and service management. Although machine learning has substantially advanced demand forecasting, existing reviews have primarily focused on forecasting methodologies and predictive performance, providing comparatively limited synthesis of how contextual information has been represented within forecasting models. This study presents a systematic literature review of machine learning-based spatio-temporal demand prediction for taxi and ride-hailing systems, following the PRISMA 2020 framework and synthesising the literature through the lens of contextual information representation. The systematic search focused primarily on studies published between 2020 and 2025 to capture recent developments in machine learning-based demand prediction. The review synthesises the evolution of forecasting objectives, contextual information representation, and machine learning paradigms. The findings demonstrate that forecasting has evolved from conventional statistical approaches towards deep learning, graph-based, and hybrid architectures, while contextual information representation has progressed more unevenly. Temporal information remains the dominant input across forecasting models, whereas richer representations of planned special events, mobility conditions, and socio-economic information remain comparatively underrepresented despite their operational importance. Based on these findings, the review develops an evidence-based taxonomy of contextual information representation and presents a conceptual synthesis linking contextual information, machine learning paradigms, forecasting tasks, and operational decision support. The review identifies richer contextual information representation particularly through event-aware and accessibility-oriented forecasting as an emerging research priority for future machine learning-based demand prediction research. |
Paper 26
Mitigator or generator? A Systematic review of how public transport shapes social segregation
Room Location: B11.B3.101
View Authors
- Shuoyu Xie (Monash University)
- Liton Kamruzzaman (Monash University)
- Alexa Delbosc (Monash University)
View Abstract
| Public transport is increasingly recognized as a contributing factor of social segregation, yet empirical evidence presents a mixed picture regarding its role. Through a PRISMA-guided systematic review of 30 studies (2008–2025), this paper synthesizes evidence on how transit interventions shape segregation outcomes along two pathways: a behavioural pathway, through which transit expands mobility and facilitates diverse encounters, and a structural pathway, through which transit-induced land-use changes reshape neighbourhood composition. Findings reveal that transit consistently mitigates activity-space segregation by expanding mobility, yet its impact on residential segregation is heterogeneous, often triggering gentrification and displacement. Importantly, gains in cross-group exposure measured by activity-space indices can come with hidden costs: displaced lower-income residents often face longer commutes and greater time-poverty, which may limit the quality of social contact even as mobility-based mixing increases. We also note that common methodological challenges—endogeneity and the modifiable area unit problem—indicate areas where future work could strengthen causal inference and use more fine-grained mobility data to track segregation patterns. The findings suggest that the social effects of transit access are shaped by operational features such as service frequency, fare structures, and network topology. Building on this synthesis, we identify priorities for future research, including longitudinal person-based tracking, stronger causal identification strategies, and richer accessibility models that capture a wider range of mobility costs. Together, these advances would help clarify the conditions under which transit investment supports more equitable urban outcomes. |
Paper 54
Car-following stability and its influence on fuel consumption and emissions: Evidence from empirical data
Room Location: B11.B3.101
View Authors
- Jiahao Zhang (The University of Queensland)
- Zuduo Zheng (The University of Queensland)
- Jiwon Kim (The University of Queensland)
- Hesham Rakha (Virgina Tech)
View Abstract
| Emission and fuel consumptions (EFCs) are strongly influenced by microscopic car-following (CF) behavior, particularly through the stability of vehicle interactions that govern speed and acceleration dynamics. However, existing studies primarily rely on aggregate traffic measures or model-based analyses, while the trajectory-level relationship between CF stability and environmental impacts remains insufficiently quantified using real-world data. This study develops a rigorous framework to investigate how driving stability affects EFCs by integrating trajectory-level stability classification with multi-model EFC comparison. Using high-resolution trajectory from the Arizona Ring Experimental Dataset (ARED), each trajectory is represented by calibrated parameters of the Fadhloun–Rakha (FR) CF model. A simulation-based string stability analysis is then applied to classify driving behavior into stable and unstable driving. Four representative EFC models (VT-CPFM, VSP, MEF, and ARRB) are employed to evaluate environmental impacts, enabling a comprehensive assessment across instantaneous and trajectory-level metrics. Results show that unstable driving significantly increases vehicle’s fuel consumption and emissions. Unstable trajectories exhibit a significantly higher frequency of extreme high-fuel-use events, along with consistently elevated fuel consumption and emission levels per unit distance across all EFC models. Therefore, the findings demonstrate that driving instability significantly increases fuel consumption and emissions, while emphasizing the importance of stability-oriented strategies and policy interventions for enhancing fuel efficiency and reducing environmental impacts in traffic systems. |
Paper 93
An agent-based simulation of bushfire evacuation: A case study of Black Summer 2019-2020 in South Coast, NSW
Room Location: B11.B3.101
View Authors
- Hitomi Nakanishi (University of Canberra)
- Nishant Jegannath (University of Canberra)
- Milica Muminovic (University of Canberra)
- Wendi Han (University of Canberra)
View Abstract
| This study simulated the evacuation behaviour of residents and tourists in South Coast, NSW at the time of Black Summer Bushfire 2019-2020. The area has limited major roads and experience severe traffic in the summer holiday season every year. At the time of bushfire, residents experienced even severe congestion, which was made worse due to closure of these roads, and made evacuation chaotic. In this study, an agent-based simulation platform was developed, by integrating Protective Action Decision Model. The platform is implemented in Python 3.9+ and organised as a five-layer architecture, with the active layers advancing in lockstep on a fixed sequence (fire → roads → agents). Fire and fire spread data was obtained from NSW National Parks and Wildlife Service and NASA’s Earth Observing System Data and Information System (EOSDIS). Three scenarios were simulated: 1) BAU representing the situation of evacuation on 31 December 2019; 2) with a new detour route to Federal Highway (M23) and 3) with more evacuation shelters. The results show that evacuation outcomes are improved (i.e. higher number of residents evacuated successfully) with scenarios 2 and 3. The evacuation completion rate is much higher with scenario 3. The paper concludes with implications to transport and urban planning, evacuation strategies and future research direction. |
Paper 144
On-road heavy-duty vehicle emissions models: A review
Room Location: B11.B3.101
View Authors
- Randy Franstiarajah (University of New South Wales)
- Elnaz Irannezhad (University of New South Wales)
- Ruth Fisher (University of New South Wales)
View Abstract
| Heavy-duty vehicles (HDVs) are major contributors to road transport emissions despite accounting for a relatively small share of the global vehicle fleet. This review study synthesises HDV emission measurement techniques, key emission-influencing factors, and existing HDV emission modelling approaches. The review study further examines the applications of HDV emission models across macro-, meso-, and micro-scale frameworks, considering their spatial and temporal resolutions for emissions inventory development. Despite their widespread adoption, significant challenges remain in model selection and emission disaggregation, primarily due to data limitations and the absence of a standardised framework for model selection. To address these challenges, future research will develop a decision-support system (DSS) to support systematic model selection based on data availability and application-specific requirements. Finally, this review study provides recommendations tailored to the Australian context to support policymakers in developing proactive strategies for HDV emission mitigation. |
Paper 19
Discovering OD flow communities in shared bike systems: A graph-mining approach with LLM-assisted interpretation
Room Location: B11.B3.103
View Authors
- Ling Wu (Xihang University)
- Jiao Yan (Xihang University)
- Man Yu (Chang'an University)
- Dan He (The University of Queensland)
- Chen Zhang (Xihang University)
View Abstract
| The rapid growth of dockless shared bike systems has generated massive origin-destination (OD) trip data, creating new opportunities to understand how urban cycling demand is organised in space and time. Existing studies, however, mainly focus on demand intensity at stations or spatial cells, which is useful for identifying hotspots but less effective for revealing coordinated flow structures between origins and destinations. Such flow-based structures are important for transport planning because they can represent functional demand corridors relevant to rebalancing, feeder-service design, and corridor-level infrastructure improvement. To address this gap, this study proposes a graph-mining framework to discover OD flow communities in shared bike systems. Trips are first aggregated into OD edges using H3 hexagonal grids, and a weighted OD-of-OD graph is constructed by integrating temporal, spatial, network, and co-occurrence similarity. Louvain community detection is then applied to identify higher-order groups of OD flows with similar mobility patterns. The framework is evaluated on a week of dockless shared bike data from Shenzhen, China, comprising approximately 10.4 million valid trips. The results show that the detected communities exhibit clear temporal and spatial differentiation, including AM-dominant, PM-oriented, and balanced weekday structures. A case study on one representative community further illustrates how LLM-assisted interpretation can help translate quantitative community profiles into concise transport-oriented semantic descriptions. Overall, the findings show that analysing OD relations, rather than locations alone, provides a more useful basis for understanding shared bike mobility organisation and for supporting planning and operational decisions. |
Paper 134
Examining modal shift under transport disadvantage conditions: Evidence from a stated choice experiment
Room Location: B11.B3.103
View Authors
- Namrata Ghosh (Queensland University of Technology)
- Dilini Kariyawasam Pathiranalage (Queensland University of Technology)
- Ishani Shehara (Queensland University of Technology)
- Shamsunnahar Yasmin (Queensland University of Technology)
View Abstract
| Transport disadvantage (TD) and modal shift (MS) are often examined as separate transport policy problems, despite their interdependence, particularly in regional and remote areas where private vehicle dependence is shaped by limited accessibility and constrained transport alternatives. This study examines behavioural mode-choice mechanisms under TD conditions and evaluates how short-term operational improvements influence predicted modal shift outcomes in Western Sydney, regional, remote, and very remote areas of New South Wales (NSW), Australia. A stated-choice experiment was conducted among residents of Western Sydney, regional, remote and very remote areas of New South Wales (NSW), Australia, yielding 936 observations from 234 respondents. A Multinomial Logit (MNL) model was estimated to analyse behavioural trade-offs associated with travel time, cost, waiting conditions, and perceived safety across four travel alternatives: private vehicle, carpool/shared ride, walking + public transport, and cycling/micromobility + public transport. Results indicate persistent structural preference toward private vehicle use, with unsafe walking environments, affordability pressures, and operational uncertainty significantly constraining non-car alternatives. Scenario analysis demonstrates that operational improvements generate only modest modal shifts even under favourable conditions, with private vehicle travel remaining dominant across all scenarios. The findings suggest that operational interventions alone may support only incremental modal shift unless broader structural accessibility gaps associated with TD are simultaneously addressed. |
Paper 149
The travel mode share and patterns of the ‘four types of commuters’ across Sydney and Brisbane in 2016
Room Location: B11.B3.103
View Authors
- David Hayward (Aptura Consulting)
- Harry Wilson (Aptura Consulting)
- Matthew Petherick (Aptura Consulting)
View Abstract
| The ‘four types of commuters’ framing of work trip patterns, established by Peter Seamer, former CEO of Victoria’s Planning Authority, classifies work trips into ‘inward’, ‘short’, ‘circumferential’ and ‘contraflow’ categories. Previous Australasian Transport Research Forum (ATRF) papers applied this framework to Greater Brisbane using ABS Census data between 2011 and 2021 (Hayward et al., 2025), revealing significant variation in travel mode shares across work trip types. This paper extends that analysis to compare Brisbane with Sydney using ABS 2016 Census Journey to Work data. For each city, the spatial distribution of commuter types is established at the Statistical Area Level 2 (SA2) and travel mode shares are calculated for each work trip category. Comparative analysis across the two cities reveals distinct patterns in car dependency and the level of public transport and active travel utilization by work trip type, reflecting differences in urban form, population density and transport network supply. The findings demonstrate consistent dominance of car travel for ‘circumferential’ and ‘contraflow’ trips across both cities, with ‘inward’ trips recording the highest public transport mode shares in each city. Significant variation between Sydney and Brisbane is identified for ‘inward’ and ‘short’ trip public transport and active travel mode shares, reflecting the differing maturity of public transport networks and urban density across the cities. These insights provide a comparative basis to inform bespoke transport planning strategies and targeted infrastructure investment for each city as populations grow towards 2046. |
Paper 14
How competitive are regional coaches with car travel to Melbourne? – the impact of the 2021 fare reforms
Room Location: B11.B3.104
View Authors
- Lachlan Wong (Monash University)
- Rumana Sarker (Monash University)
- Graham Currie (Monash University)
View Abstract
| Due to a high share of regional Victorians relying on coach services for access to central Melbourne, there is a need to explore how well these settlements are served. Using a Total Generalised Cost (TGC) modelling, this paper develops a spatial understanding of how accessible and competitive regional V/Line coach services are compared to private transport from outer Victoria to Melbourne’s CBD. Results show that regional coach services have become more cost-competitive following the fare reform in 2021, which capped fares at $11.00. From a journey time perspective, on average, coaches took 48-minutes longer than private transport. However, from a generalised cost perspective, coach journey cost was lower than car for both the perceived and total car costs. The fare change reduced the total generalised cost of coach journeys by $13.62 for morning peak arrival (9am arrival times) and $18.43 for afternoon peak arrival (2pm arrival times); with the largest benefits observed for longer-distance regional trips. These findings highlight the role of fare policy in improving affordability and competitiveness, particularly in the context of rising fuel prices and broader cost-of-living pressures, while also underscoring the importance of addressing travel time penalties to support mode shift. |
Paper 47
Built by rail, shaped by time? The link between train stations and population density
Room Location: B11.B3.104
View Authors
- Qinran Zhang (Monash University)
- Liton Kamruzzaman (Monash University)
- Graham Currie (Monash University)
View Abstract
| Railway infrastructure is widely associated with urban densification, yet the causal effect of contemporary train station proximity on population density remains difficult to identify because station locations are often endogenous to urban development. Focusing on the Greater Melbourne Area, this study examines whether proximity to train stations has an independent effect on contemporary population density. Residential population density is measured at the Statistical Area Level 1 using 2021 Census data, and rail accessibility is captured by the walkable network distance from each SA1 centroid to the nearest train station. To address endogeneity, contemporary station distance is instrumented by the Euclidean distance to the nearest station in Melbourne’s 1870 railway network, drawing on the long-run persistence of rail corridor development. The two-stage least squares results show a robust and statistically significant density gradient. In the preferred controlled specification, a one-kilometre increase in distance to the nearest train station is associated with about 827 fewer persons per square kilometre. The IV estimates are substantially larger than the corresponding ordinary least squares estimates, indicating that conventional correlation-based models understate the density effect of rail accessibility. The findings suggest that train stations are not only located in denser areas, but also play an independent role in shaping where the population concentrates over time. |
Paper 56
Human-AI Collaboration (HAIC)-enabled retrieval-augmented generation (RAG) framework for railway accident investigation
Room Location: B11.B3.104
View Authors
- Fei-Fan Wen (National Yang Ming Chiao Tung University)
- Wei-Ting Hong (The University of Western Australia)
- Chen-Yu Lin (National Yang Ming Chiao Tung University)
View Abstract
| Railway accident investigation relies heavily on expert judgement and manual synthesis of extensive unstructured reports, limiting scalability and cross-jurisdictional learning. Based on the HazardMap framework, this study proposes a Human-AI Collaboration (HAIC)-enabled retrieval-augmented generation (RAG) framework for railway accident investigation. Using 421 reports from the Rail Accident Investigation Branch (RAIB) and 28 reports from the Taiwan Transportation Safety Board (TTSB), BERTopic and graph-based knowledge representation are integrated with a LightRAG system to support evidence-based reasoning. The results demonstrate how recommendation patterns differ across jurisdictions and identify potential railway safety knowledge gaps. A case study of the Taichung MRT crane collapse accident further illustrates that the system can generate traceable causal reasoning and recommendations comparable to expert investigation outcomes. The study presents an initial proof-of-concept demonstrating the feasibility of operationalising structured hazard knowledge to support HAIC in railway safety decision-making. |
Paper 137
Bustrify: a simulation and analysis package for electric public bus fleets
Room Location: B11.B3.104
View Authors
- Ahmed Abdelhadi (University of Auckland)
- Minh Kieu (University of Auckland)
- Andrea Raith (University of Auckland)
View Abstract
| The electrification of public bus fleets offers clear environmental benefits, but it also introduces operational, planning, and infrastructure challenges that extend well beyond vehicle replacement. Compared with diesel buses, electric buses face constraints associated with limited driving range, non-trivial charging times, and tighter coupling between vehicle availability, route structure, and depot operations. These characteristics can reduce operational flexibility, particularly in networks designed around interlining and high daily utilisation, and may require changes to fleet sizing, route assignment, or charging infrastructure deployment. As a result, electrification decisions increasingly demand system-level analysis tools capable of capturing time-dependent interactions across vehicles, schedules, and energy systems. This paper presents Bustrify, a discrete-event simulation package developed to support the analysis and planning of electric bus operations under realistic, system-scale constraints. Bustrify explicitly models bus movements, service schedules, battery state-of-charge evolution, and charging infrastructure utilisation, enabling assessment of operational feasibility, emergent bottlenecks, and trade-offs between vehicle and infrastructure technologies and design choices. The primary contribution is the simulation package itself, intended as a transparent and extensible tool for evaluating electrification strategies. To demonstrate the capabilities of Bustrify, we explore a case study of the inner Auckland bus network using real route and timetable data. The case study explores battery-sizing trade-offs in a fully electrified fleet supported by static wireless charging, which enables opportunity charging during brief dwell periods. To assess these trade-offs, battery capacity is progressively reduced from a 400 kWh baseline while service schedules are held fixed. The results show that wireless charging enables substantial battery downsizing without immediate loss of service reliability, with operations remaining feasible when battery capacity is more than halved under the assumed infrastructure configuration. Bustrify thereby allows us to identify a lower bound on battery capacity beyond which operations degrade significantly. Overall, the case study illustrates how Bustrify can be used to reveal non-linear, network-level effects of electrification decisions and support evidence-based planning. The package is intended for open-source release to facilitate reproducible research and broader application. |
3:10 – 3:30
Afternoon Tea
3:30 – 5:00
Plenary 4: Closing & Awards
Keynote speaker:
-
Professor Kevin J. Krizek, Professor of Transport in the Programs of Environmental Design and Environmental Studies at the University of Colorado Boulder.
UTS Great Hall