Summary: UCLA researchers in the Departments of Civil & Environmental Engineering and Computer Science have developed a highly scalable generative deep learning framework that eliminates the high costs associated with traditional human mobility modeling.
Background: Consistent and accurate human mobility modeling is crucial for urban planning, transit development, infrastructure optimization, and regional economic analysis. Traditional human mobility modeling has relied on activity-based models (ABMs) that predict how people move through a geographic area. However, these traditional models rely on rigid behavioral assumptions and require large, expensive datasets to function. Because of this data dependency, traditional ABMs are costly and difficult to adapt to new regions, and this issue is exacerbated in developing areas and municipalities with limited data. In order to modernize regional planning, there is an unmet need for a flexible, data-efficient modeling solution that can accurately synthesize human mobility in a scalable and cost-effective manner.
Innovation: Researchers at UCLA have developed a novel generative deep learning framework for human mobility modeling and synthesis that captures complex population movements using widely available, open-source data. Unlike traditional approaches, this technology simultaneously incorporates high-level activity patterns and precise location trajectories to simulate realistic regional movement. The deep learning model can be efficiently fine-tuned with minimal data, allowing it to rapidly adapt to unique mobility fingerprints of diverse populations. Additionally, the framework incorporates advanced data-fusion methods capable of blending multimodal travel datasets, enabling a highly comprehensive and unified view of regional transit dynamics. By resolving the rigidity and high cost associated with traditional ABM, this software framework enables the widespread adoption of advanced mobility analytics for urban planners and transit authorities globally.
Potential Applications: ● Urban planning & infrastructure development ● Public transit optimization ● Smart city & IoT analytics ● Geospatial site selection
Advantages: ● High data efficiency ○ Low cost ● Regional adaptability ● Unified behavior synthesis ● Multimodal integration ○ Data-fusion capabilities for transit and travel data
Development-To-Date: The framework is currently being positioned for real-world pilot testing and collaborative validation with early partners. Customer interest garnered
Reference: UCLA Case No. 2026-199
Lead Inventor: Jiaqi Ma, Professor of Civil & Environmental Engineering, Professor of Computer Science