New GeoAI roadmap calls for equity-focused AI in global transportation systems

Urban mobility systems worldwide are failing to deliver equitable access, and new research suggests artificial intelligence may hold both the diagnosis and the cure. A new review highlights how emerging GeoAI technologies, when combined with multimodal geospatial data, are reshaping the way cities understand and address mobility inequality. The study argues that transportation is no longer just a matter of infrastructure but a structural determinant of social inclusion, economic opportunity, and environmental justice.

Published in Urban Science, the study titled “GeoAI and Multimodal Geospatial Data Fusion for Inclusive Urban Mobility: Methods, Applications, and Future Directions” finds that integrating artificial intelligence with diverse geospatial datasets, ranging from satellite imagery and GPS traces to social media and participatory mapping, allows cities to uncover hidden mobility gaps with unprecedented precision.

AI-powered mobility analysis exposes deep-rooted urban inequalities

The research makes clear that mobility inequality is both widespread and systemic. Across cities in the United States, Europe, and Latin America, disadvantaged populations consistently face reduced access to public transit, micromobility options, and essential services. In some cases, nearly 80 percent of residents in marginalized neighborhoods lack access to bikes and scooters, while low-income commuters endure significantly longer travel times due to poorly located transit infrastructure.

These disparities are not isolated inefficiencies but reflections of broader structural inequalities tied to income, race, geography, and urban planning decisions. The study frames mobility as a core element of spatial justice, noting that limited transportation access restricts employment opportunities, healthcare access, and education outcomes. It also exposes vulnerable populations to higher environmental risks such as traffic pollution and unsafe street conditions.

GeoAI, a field that combines machine learning with geographic information science, is emerging as a powerful tool to address these challenges. Unlike traditional planning methods that rely on periodic surveys and aggregated statistics, GeoAI systems integrate real-time, high-resolution data from multiple sources. These include satellite imagery, sensor networks, GPS mobility data, and user-generated geographic information.

By fusing these datasets, AI models can map accessibility gaps, predict demand patterns, and identify underserved communities with far greater accuracy than legacy systems. Deep learning models such as convolutional neural networks and graph neural networks enable detailed analysis of urban environments, from detecting sidewalk barriers to modeling transit network inefficiencies.

However, the study also warns that these technologies can replicate existing inequalities if not carefully designed. Many datasets, such as GPS traces and social media activity, disproportionately represent affluent users, leading to biased models that overlook marginalized populations. As a result, the authors emphasize the need for fairness-aware algorithms and inclusive data collection practices.

Multimodal data fusion transforms how cities plan and predict mobility

A key contribution of the study lies in its analysis of multimodal geospatial data fusion, a process that combines different types of data to create a more complete picture of urban mobility systems. The research identifies three main approaches: data-level fusion, feature-level fusion, and decision-level fusion.

Feature-level fusion emerges as the most widely used method, allowing AI models to integrate diverse data streams such as weather conditions, infrastructure attributes, and mobility patterns into unified representations. This approach improves predictive accuracy and enables more nuanced insights into how urban systems function.

The applications of these techniques are wide-ranging. In accessibility mapping, AI models can combine street-level imagery with demographic data to identify neighborhoods where vulnerable groups face compounded mobility barriers. In demand forecasting, multimodal systems integrate transit usage data, mobile phone records, and environmental factors to predict travel patterns and optimize services.

The study finds that most existing GeoAI applications focus heavily on predictive tasks such as traffic forecasting and origin-destination modeling. While these models achieve strong performance improvements, they often prioritize efficiency over equity. Only a limited number of studies explicitly incorporate fairness metrics or evaluate how predictions vary across different population groups.

Geographic coverage also remains uneven. Most research is concentrated in data-rich cities in China and the United States, leaving large parts of the Global South underrepresented. This raises concerns about the transferability of AI models, as systems trained in highly digitized environments may fail in contexts with informal transport systems and limited data availability.

Despite these limitations, the potential of GeoAI for urban planning is significant. The study highlights how these tools can help identify “mobility deserts,” evaluate policy scenarios, and guide infrastructure investments toward underserved communities. Real-time data integration also enables continuous monitoring of transportation systems, allowing cities to respond more quickly to disruptions and changing demand patterns.

Governance, ethics, and participation emerge as critical challenges

While the technological capabilities of GeoAI are advancing rapidly, the study stresses that technical innovation alone is insufficient to achieve equitable mobility outcomes. Ethical, governance, and institutional challenges remain major barriers to implementation.

One of the most pressing issues is data bias. Crowdsourced platforms and digital data sources often underrepresent low-income populations, older adults, and individuals without access to technology. This can lead to models that systematically overlook the needs of the most vulnerable groups.

Privacy concerns also loom large. Mobility data, particularly GPS traces and mobile phone records, can reveal sensitive information about individuals’ movements and behaviors. While techniques such as federated learning offer privacy-preserving solutions, governance frameworks for data sharing and protection are still evolving.

Another critical gap lies in validation practices. The study finds that most GeoAI models are evaluated using technical performance metrics such as accuracy or error rates, with little attention to equity outcomes. Models that perform well on average may still produce biased results for specific communities, masking underlying disparities.

To address these challenges, the authors call for a shift toward participatory and human-in-the-loop approaches. Incorporating community input through participatory mapping and stakeholder engagement can help ensure that AI-driven insights align with lived experiences. This is particularly important in contexts where local knowledge about safety, accessibility, and cultural factors may not be captured in data.

The study also calls for new governance models, such as urban data trusts, to facilitate ethical data sharing and cross-sector collaboration. Transparent workflows, open data practices, and reproducibility standards are essential to building trust in GeoAI systems and ensuring their long-term impact.