End-to-End deep learning for inventory management in capacitated bike-sharing systems with customer roaming behaviour

T-CT2CRP-SM: The exploration of dockless bike-sharing system rebalancing problems based on multi-modal signals in social networks

ABSTRACT Urban cycling plays a vital role in public health promotion and sustainable mobility, yet high summer temperatures often discourage cycling. Street shading can mitigate heat stress and enhance outdoor thermal comfort, but its effectiveness in promoting summer cycling remains unclear. This study leverages multi-source data–including Baidu street-view imagery and bike-sharing records–and employs XGBoost combined with SHAP analysis to investigate the relationship between street shading (tree shading and building shading) and residents’ cycling behavior in Shanghai. Our results indicate that: (1) shared bicycle usage follows a ‘dense-center, sparse-periphery’ pattern, with substantial unevenness in street shading across central districts; (2) street shading can significantly increase cycling frequency, with building shading exerting a stronger effect than street-tree shading; and (3) both building and street-tree shading display nonlinear effects, markedly boosting cycling frequency once certain thresholds are exceeded, though excessive building shading produces diminishing returns. These findings suggest that strategically positioned and appropriately scaled street and building shading can improve thermal comfort, encourage active mobility, and support sustainable cycling in dense urban environments.

Bicycle-sharing systems (BSS) have become an important component of sustainable urban mobility, but their demand remains difficult to model. Usage varies across hours, stations, weather conditions, and types of day, while the available data often provide only a partial view of the underlying demand process. This study proposes an interpretable probabilistic framework to characterize and generate synthetic demand for BiciMad, Madrid's dock-based BSS, using trip-level data from 2018 and 2019. The contribution is not the introduction of new probability distributions, but the calibrated integration of standard probabilistic components into a demand-side generative framework. Trip distances are modeled with Gamma distributions, and hourly trip counts are represented with Negative Binomial distributions conditioned on hour, day type, and precipitation. These components are combined with empirical station-popularity profiles to generate synthetic origin-destination demand under explicit contextual assumptions. Validation against observed data shows that the framework provides calibrated uncertainty estimates, with empirical coverage of the 95% prediction intervals close to the nominal level across contextual scenarios. An external consistency check using 2019 data further shows the practical value of the approach, as it helped identify systematic timestamp misattributions that were later confirmed by the data provider. The proposed framework is not intended as a full capacity-aware operational simulator. Instead, it provides a simple, interpretable, and uncertainty-aware baseline for demand characterization, synthetic demand generation, exploratory disruption analysis, and data-quality consistency checking in dock-based bicycle-sharing systems.

Accurate demand forecasting is critical for optimising resource allocation and operational efficiency in bike-sharing systems (BSS). However, this remains challenging owing to the complex nonlinear characteristics of large-scale data, irregular spatiotemporal patterns, and external environmental uncertainties. To address these issues, this study proposes a multidimensional spatiotemporal fusion framework that combines Domain-Adaptive Density Clustering (DADC) and an Attention Temporal Graph Convolutional Network (A3T-GCN). First, the DADC algorithm addresses data sparsity by adaptively clustering stations based on spatial density variations, thereby effectively capturing dynamic distribution patterns. Second, a weighted directed graph is constructed incorporating station utility and full-load-rate functions to eliminate low-performance nodes. A dynamic graph sequence model is then developed to characterise the temporal evolution of spatial dependencies and enhance peak-period responsiveness. The A3T-GCN architecture integrates station-specific attributes with environmental factors through attention mechanisms, enabling the focused analysis of critical demand influencers while processing multidimensional spatiotemporal dependencies. Experimental validation using Citi Bike datasets demonstrates superior performance over benchmark models, reducing prediction errors by 11.54 % (15 min) and 12.33 % (60 min) through improved short-term accuracy and long-term stability. The framework effectively addresses the uncertainty in demand forecasting, providing methodological support for sustainable BSS operations through enhanced spatiotemporal pattern recognition and environmental adaptability.

The micromobility mindset: Socio-technical drivers of bike share scheme adoption in the UK

Adding bike simulation capacity to an activity–based travel demand model and testing policy scenarios

Joint design of transit and bike-sharing systems by multi-objective optimization considering stochastic user equilibrium

Wheels within reach: Evaluating bike share equity impacts of a pseudo-free student pass using network-distance-based kernel density

Assessing spatial and temporal transferability in models of cycleway impacts can inform evidence-based transport policies and wider implementation of cycling infrastructure. Yet, limited research on model transferability constrains their practical application in infrastructure planning. This study examines spatial and temporal transferability of models estimating the effects of cycleway implementation on bike-share usage. Using 13 years of fortnightly data from London's bike-share scheme across Cycleways 1, 3, and 6 (over 14,000 time-station observations), we combine three modelling frameworks (Autoregressive Integrated Moving Average with Exogenous variables [ARIMAX], Generalised Additive Models [GAM], and Generalised Additive Mixed Models with ARMA errors [GAMM+ARMA]) with three transferability strategies (direct transferability, contextual calibration, and local re-estimation). Results show that contextual calibration generally outperforms the other two strategies, reducing errors by 20-80% compared with direct transferability and by 5-45% compared with local re-estimation. Cycleway interventions are positively associated with bike-share usage, whilst demographic covariates exhibit spatial heterogeneity. These findings highlight the value of partial model adaptation for balancing transferability with local relevance and suggest contextual calibration as a practical strategy for transferable cycleway intervention modelling. This study provides insights for transport authorities to prioritise investment, scale cycling infrastructure efficiently, and adapt successful interventions to diverse urban contexts.

Life cycle and net environmental impact assessment of a shared e-bike service: application to the case of Madrid

The integration of cycling and rail transport is a key element of sustainable commuting, yet its success depends on the accessibility of stations, their surrounding environments, and destinations. While quantitative studies have extensively analysed infrastructural and behavioural determinants of bike–train use, fewer have examined how users perceive accessibility in practice. This article explores how accessibility conditions shape the bike–train commuting experience in Flanders, Belgium. Drawing on five focus groups with university staff across four campuses with varying levels of bike–train accessibility, the analysis identifies critical accessibility dimensions at both the home-end and activity-end of trips. Physical accessibility factors, including bicycle parking availability, visibility and proximity to platforms, strongly shape users’ evaluations of station quality. Accessibility is further influenced by the presence of cycle highways, multiple station entrances, and the availability and coverage of shared bicycle systems. Beyond physical elements, perceived accessibility is affected by train frequency and reliability, ease of use, perceived safety, legibility and design of the public realm, and the atmosphere and services provided at stations. The findings demonstrate that accessibility challenges and expectations differ substantially between home-end and activity-end stations: the former is primarily evaluated based on functional access to platforms, while the latter is judged more broadly on service reliability and the comfort of the station environment. While limited to current users, this study offers qualitative evidence on how accessibility conditions shape the usability of bike–train commuting, informing policies aimed at strengthening multimodal integration.

As a key link between short-distance urban mobility and public transport, dockless bike-sharing (DBS) systems have expanded rapidly in recent years. However, existing studies are limited by insufficient factor coverage, incomplete temporal analysis, and inadequate assessment of spatial-scale effects. To address these gaps, this study uses Shenzhen as a case study, integrating 40 million DBS trip records from August 2021 with multi-source geospatial data to develop a spatiotemporal analytical framework. First, it examines differences in riding patterns between weekdays and weekends, further segmenting trips into six time periods to capture intra-day temporal variations. Through multicollinearity and spatial autocorrelation tests, a 700-m grid was identified as the optimal analysis unit. Subsequently, a Multi-scale Geographically Weighted Regression (MGWR) model quantified how multiple sources of factors collectively shape DBS usage behavior. Results indicate that higher frequency, faster speeds, and longer distances during peak periods characterize weekday trips. Office POIs and transit accessibility positively affect DBS usage during weekday peaks, whereas Residential POIs and Convenience Service POIs have a greater influence on weekend trips. Population density and land-use mix consistently promote DBS use across all periods. Younger residents (<30 years) were the main users, especially during weekday peak and weekend no-peak periods, whereas gender and education had limited impact. These findings provide empirical evidence to optimize bike-sharing deployment, enhance multimodal transport integration, and support sustainable urban mobility planning.

In many cities worldwide, cycling is gaining popularity as a mode of transportation, offering environmental, health, and financial benefits. Yet research suggests that bicycle theft continues to deter individuals from cycling, limiting the broader adoption of this sustainable form of transportation. Although a growing body of criminological research has examined how features of the built and social environment influences crime patterns, the role of topography remains largely overlooked—particularly concerning bicycle theft. To date, few studies have explored how slope and hilliness relate to criminal behavior. This study contributes to this emerging area by applying environmental criminological theories to examine the spatial distribution of bicycle thefts in San Francisco, California, from 2018 to 2023 (n = 2,762 bicycle thefts). Various topographical measures, along with spatial access measures of built environment features, street centrality, population estimates, and social disorganization indices were included in the final negative binomial regression models. Findings show that street slope was negatively associated with bicycle theft, with streets with steeper slope gradients experiencing fewer thefts. Multiple infrastructure-related variables were significant predictors of bicycle theft, including bicycle racks, transit stops, and bike share stations while street centrality was not associated with theft. The theoretical implications of this study suggest that criminological research should incorporate measures of topography as these factors influence crime patterns by shaping offender mobility, target availability, and escape feasibility. Integrating these insights into bicycle theft prevention may help promote cycling as a sustainable mode of urban transportation.

Abstract ➖ This paper introduces a dynamic pricing-based public bike pooling platform tailored for college student transportation. Unlike existing ride-sharing platforms, which rely on static pricing models, our system adjusts pricing dynamically based on demand, supply, and time of day, ensuring equitable access and optimized resource distribution. The platform addresses the unique needs of students navigating campus environments, offering affordable and flexible ride options. Our approach integrates user behavior analysis and historical trip patterns to predict demand fluctuations. This predictive capability allows the system to encourage pooling at peak times and ensures bike availability during off-peak periods. By doing so, our platform not only enhances student mobility but also reduces operational inefficiencies. To validate our model, we conducted a study using simulated campus trip data. Results demonstrate improved cost-efficiency and increased bike utilization compared to static pricing models. In summary, this dynamic platform fills a gap in current bike-sharing services, providing a flexible and fair solution for college campuses. Our work sets the foundation for smarter, demand-responsive transportation in student communities. Keywords Bike Pooling, Dynamic Pricing, Smart Transportation, Student Mobility, Ride Sharing Systems, Demand Prediction, Cost Optimization, Sustainable Transportation, Campus Transportation, Data-Driven Systems

Trip-level heat exposure risk assessment framework for bike-sharing systems: A Case Study of Shenzhen

This paper compares trends in cycling levels, cyclist demographics and cycling injury risk in New York, London, Paris and Berlin, before and after the COVID pandemic. We explore these trends in the context of changes to policy and infrastructure before, during, and after COVID. We based our analysis on data from published reports, open-data portals, government websites, travel surveys, and information provided by transport planners in each city. Cycling levels in NYC, London, Paris, and Berlin increased over the three decades prior to COVID (1990–2019). As a percentage of daily trips, bike mode share rose from 0.6% to 2.2% in New York, from 1.2% to 3.7% in London, from 0.4% to 5% in Paris, and from 7% to 18% in Berlin. Cycling rates have continued to increase since COVID. By 2023, bike mode shares had risen further to 3% in NYC, 4.5% in London, 11% in Paris, and 19% in Berlin. Cycling became safer in all four cities over the period 2005 to 2023, with declining per-trip fatality and injury rates. More and better cycling infrastructure has been a cornerstone of pro-cycling efforts, especially cycleways separated from motor vehicle traffic (protected bike lanes). Bike parking and bikesharing systems have expanded and improved. Car restrictions and traffic calming have complemented pro-bike measures, for example, using infrastructure and enforcement to reduce traffic volumes and speeds in residential neighborhoods. Long-term political support as well as cycling advocacy organizations have been critical to the introduction and continuation of pro-bike policies and the necessary financial investments.

This study examines the Public Bicycle Sharing System (PBSS) in Ahmedabad, the patterns of ridership, user behavior, and spatial distribution in various density zones based on the data provided by MYBYK. In 2023, 577,228 rides were registered, and high ridership areas were concentered around educational and recreational locations, which recorded 350,000 rides in 132 stations. In the high-density zones, the analysis indicates that most of the preference is for the short-distance ride, 55 percent of the total number of trips are made by zero-displacement rides, which is a good indicator that the trend is shifting towards leisure cycling. The monthly ridership trends show some significant variations, with the highest number of 68,529 rides in March, indicating that significant operation policies should be implemented when demand is very high. The dominance of short-term subscriptions (87.3%) indicates the users' preference for flexibility, whereas the number of rides longer than 5 km (424,893 trips) shows a growing tendency towards longer trips. Furthermore, a multinomial logistic regression model was used to identify factors affecting the use of PBSS, with emphasis on the relationships between the type of subscription, length of trip, time of use, and weather conditions. The model findings indicate that short-term users tend to have shorter rides, and the length of the trip is highly influenced by climatic conditions. The results can be useful to policymakers and urban planners to optimize the PBSS operations and improve sustainable mobility in Ahmedabad and other urban settings.

Pedal preferences: GPS-based panel data insights into bike share traffic flow across membership groups

The built environment supports physical activity (PA) by providing opportunities to be active in daily life. Natural experiments are valuable for assessing how real-world changes to the built environment affect PA and are critical for guiding policies to improve population-level PA. The objective of this review was to summarize the evidence from natural experiments that investigated the impacts of built environment changes on PA in Canada. Searches were conducted in MEDLINE, Embase, PsycINFO, ProQuest Public Health and SportDISCUS, from inception to 27 November 2024. Natural experiment evaluations that included a comparator or historical control group and assessed changes in PA associated with changes in the built environment were eligible. A narrative synthesis summarizes the evidence and the certainty of the evidence. Results from the included natural experiments (n = 25) suggest positive effects, with low to moderate certainty, of increased walkability, new cycling and pedestrian infrastructure, bike share (bike rental) programs and new trails. However, there was very low to low certainty of no significant effects for bus rapid transit, school building and yard improvements and school zone improvements. Some evidence suggests negative effects of off-leash dog park areas on children's park-based PA and of daycare yard improvements on moderate-to-vigorous intensity PA. Few Canadian studies have evaluated the impact of built environment changes on PA, with most emerging in the last decade. Future studies should include larger and more diverse samples and all regions, control for confounders including seasonal variation in outdoor PA, use well-matched control groups and incorporate objective PA measures.