Presentations and Posters

Welcome to the Presentations and Posters page of the Big Data, AI and Transportation Planning Applications Workshop. This section highlights the cutting-edge research and innovative solutions presented by our esteemed speakers and contributors.

Here, you will find detailed abstracts and insights into the diverse topics covered during the workshop, from data-driven studies and modeling approaches to the latest advancements in AI and transportation technologies. Explore the presentations and posters to deepen your understanding of the challenges and opportunities in transportation planning, and how big data and AI are driving transformative changes in the field.

> > > Workshop Presentations

Speaker	Title	Abstract
Dr. Yueyue Fan (UC-Davis)	Where does domain expertise stand in transportation AI research? (Recording) (Slides)	In this presentation, Dr. Fan will discuss the roles of transportation domain expertise in an era of fast growing AI technologies. Dr. Fan will use examples to demonstrate how domain knowledge, when incorporated across steps like data selection and model design, may help improve the effectiveness and interpretability of AI models. Dr. Fan will also discuss barriers that need community-level efforts to foster collaboration between domain experts and AI practitioners. The goal is to ignite a conversation about how domain expertise can drive innovation and enhance the impact of AI in solving pressing transportation challenges.
Dr. Prateek Bansal (NUS)	Harnessing Household Travel Survey with Passively Collected Mobility Data (Recording) (Slides)	Household Travel Survey (HTS) data is commonly used in activity-based models but faces limitations such as infrequent collection and low spatial heterogeneity. Passively Collected Mobility (PCM) data addresses these shortcomings by offering greater spatial and temporal coverage but lacks several key attributes. This talk discusses analytical non-parametric approaches to integrate HTS and PCM data, ensuring that the rich distributions of variables from each dataset are preserved in the fused representation of sociodemographic profiles and daily activity schedules. The case studies will demonstrate how spatiotemporally heterogeneous digital mobility twins can be generated by harnessing HTS data with cellular trace data and transit smart card data for Singapore and Seoul, South Korea.
Dr. Cynthia Chen (UW)	Biases in the big data and what we can do to unleash its potential (Recording) (Slides)	This talk presents an overview of both traditional (household travel surveys) and emerging data sources (location-based-service data) used for transportation applications. In particular, Dr. Chen will present biases identified on the LBS data as well as our thoughts for future directions for the field. The goal is to provoke thoughts and discussions within the community.
Dr. Yan Liu (USC)	A New Era Dawns: AI for New Generation Transportation System (Recording) (Slides)	Recent development in deep learning has spurred research advances in time series modeling and analysis. Practical applications of time series raise a series of new challenges in transportation applications, such as multi-resolution, multimodal, missing value, and interpretability. In this talk, Dr. Liu will discuss possible paths to foundation models for time series data and future directions for time series research.
Dr. Carolina Osorio (Google Research, HEC Montréal)	Combining physics-informed machine learning with transportation science methods to tackle large-scale urban mobility problems (Recording) (Slides)	This talk presents various physics-informed methods to search high-dimensional continuous spaces in a sample efficient way, with a focus on urban mobility applications. Advances in three areas will be presented: (i) variance reduction methods for gradient estimation, (ii) sample-efficient dimensionality reduction methods, (iii) sample-efficient simulation-based network optimization algorithms. Dr. Osorio will present case studies based on various metropolitan areas, and identify and discuss research opportunities and challenges in the fields of optimization and machine learning as applied to urban mobility problems.
Dr. Sean Qian (CMU)	Data-friendly network modeling through computational graphs: learning, prediction, and decision making (Recording) (Slides)	With the availability of various data sources across all modes of transportation systems, it remains a challenge how to take advantage of those diverse spatio-temporal data to best understand travel patterns across those modes in high spatio-temporal resolutions. In a mesoscopic network modeling framework, Dr. Qian formulate and solve for spatio-temporal passenger and vehicular flows in a multi-modal network explicitly considering solo-driving, public transit, parking, curb use and ride-sharing. Through a computational graph approach, the travel behavior models and network characteristics can be jointly learned from a generic set of data, e.g. time-varying counts, speeds, census, transit data, and curb use data. Machine learning (ML) techniques are employed to optimally tune generic parameters to fit the multi-source data. This framework has been applied in many use cases for regions, cities and communities to make optimal decisions in transportation planning. The mesoscopic modeling approach can also be applied to real-time traffic operations, particularly early anomaly detection and proactive traffic management.
Dr. Takahiro Yabe (NYU)	Resilience of urban socioeconomic networks to behavioral changes (Recording) (Slides)	Urban economic resilience hinges on understanding how shocks propagate across local businesses and amenities during pandemics, disasters, and technological shifts. While disruptions in supply chains have been extensively studied, it is imperative to recognize that human behavior changes may also amplify shocks to businesses and amenities that are connected via mobility and lifestyle patterns. In this talk, Dr. Yabe will present data-driven spatial network models to predict the cascades of shocks across cities, and further discuss applications to optimize civil infrastructure systems to achieve urban resilience. Dr. Yabe will also discuss his ongoing research on cross-city transfer learning approaches to prepare cities for unprecedented shocks.
Dr. Yafeng Yin (UMich)	LLM-Agent-Based Simulation for Travel Demand Forecasting (Recording) (Slides)	This talk explores the potential of leveraging LLM agents to advance activity-based modeling of travel demand. Traditional agent-based simulations rely on hardcoded condition-action rules for agents to schedule daily activities and make travel plans. These agents adapt and interact within dynamic environments but are constrained by predefined behavioral assumptions. Dr. Yin proposes replacing these agents with LLM agents that use large language models as reasoning engines. Unlike traditional agents, LLM agents derive decision rules from the parametric knowledge embedded in the LLM, eliminating the need for manual rule-coding. This paradigm shift could enhance activity-based microsimulation by relaxing behavioral assumptions, reducing data requirements, and offering greater flexibility to evaluate innovative plans and policies.
Dr. Xilei Zhao (UF)	Enhancing Travel Behavior Analysis with Generative AI: Integrating Survey and Location Data (Recording) (Slides)	Household travel survey data have long been essential for analyzing travel behavior. While valuable, these data face major limitations, including high data collection costs, small sample sizes, and insufficient spatial details about trips. Mobile device location data, on the other hand, have recently gained attention in transportation planning. They provide large-scale, granular travel trajectories at a low cost but lack critical traveler information, such as sociodemographics and preferences/attitudes. This talk introduces a novel Generative AI framework that integrates household travel survey data with mobile device location data. By combining their strengths, this approach addresses existing limitations and paves the way for next-generation transportation planning.

Speaker

Title

Abstract

Dr. Yueyue Fan

(UC-Davis)

Where does domain expertise stand in transportation AI research?

(Recording) (Slides)

In this presentation, Dr. Fan will discuss the roles of transportation domain expertise in an era of fast growing AI technologies. Dr. Fan will use examples to demonstrate how domain knowledge, when incorporated across steps like data selection and model design, may help improve the effectiveness and interpretability of AI models. Dr. Fan will also discuss barriers that need community-level efforts to foster collaboration between domain experts and AI practitioners. The goal is to ignite a conversation about how domain expertise can drive innovation and enhance the impact of AI in solving pressing transportation challenges.

Dr. Prateek Bansal

(NUS)

Harnessing Household Travel Survey with Passively Collected Mobility Data

(Recording) (Slides)

Household Travel Survey (HTS) data is commonly used in activity-based models but faces limitations such as infrequent collection and low spatial heterogeneity. Passively Collected Mobility (PCM) data addresses these shortcomings by offering greater spatial and temporal coverage but lacks several key attributes. This talk discusses analytical non-parametric approaches to integrate HTS and PCM data, ensuring that the rich distributions of variables from each dataset are preserved in the fused representation of sociodemographic profiles and daily activity schedules. The case studies will demonstrate how spatiotemporally heterogeneous digital mobility twins can be generated by harnessing HTS data with cellular trace data and transit smart card data for Singapore and Seoul, South Korea.

Dr. Cynthia Chen

(UW)

Biases in the big data and what we can do to unleash its potential

(Recording) (Slides)

This talk presents an overview of both traditional (household travel surveys) and emerging data sources (location-based-service data) used for transportation applications. In particular, Dr. Chen will present biases identified on the LBS data as well as our thoughts for future directions for the field. The goal is to provoke thoughts and discussions within the community.

Dr. Yan Liu

(USC)

A New Era Dawns: AI for New Generation Transportation System

(Recording) (Slides)

Recent development in deep learning has spurred research advances in time series modeling and analysis. Practical applications of time series raise a series of new challenges in transportation applications, such as multi-resolution, multimodal, missing value, and interpretability. In this talk, Dr. Liu will discuss possible paths to foundation models for time series data and future directions for time series research.

Dr. Carolina Osorio

(Google Research, HEC Montréal)

Combining physics-informed machine learning with transportation science methods to tackle large-scale urban mobility problems

(Recording) (Slides)

This talk presents various physics-informed methods to search high-dimensional continuous spaces in a sample efficient way, with a focus on urban mobility applications. Advances in three areas will be presented: (i) variance reduction methods for gradient estimation, (ii) sample-efficient dimensionality reduction methods, (iii) sample-efficient simulation-based network optimization algorithms. Dr. Osorio will present case studies based on various metropolitan areas, and identify and discuss research opportunities and challenges in the fields of optimization and machine learning as applied to urban mobility problems.

Dr. Sean Qian

(CMU)

Data-friendly network modeling through computational graphs: learning, prediction, and decision making

(Recording) (Slides)

With the availability of various data sources across all modes of transportation systems, it remains a challenge how to take advantage of those diverse spatio-temporal data to best understand travel patterns across those modes in high spatio-temporal resolutions. In a mesoscopic network modeling framework, Dr. Qian formulate and solve for spatio-temporal passenger and vehicular flows in a multi-modal network explicitly considering solo-driving, public transit, parking, curb use and ride-sharing. Through a computational graph approach, the travel behavior models and network characteristics can be jointly learned from a generic set of data, e.g. time-varying counts, speeds, census, transit data, and curb use data. Machine learning (ML) techniques are employed to optimally tune generic parameters to fit the multi-source data. This framework has been applied in many use cases for regions, cities and communities to make optimal decisions in transportation planning. The mesoscopic modeling approach can also be applied to real-time traffic operations, particularly early anomaly detection and proactive traffic management.

Dr. Takahiro Yabe

(NYU)

Resilience of urban socioeconomic networks to behavioral changes

(Recording) (Slides)

Urban economic resilience hinges on understanding how shocks propagate across local businesses and amenities during pandemics, disasters, and technological shifts. While disruptions in supply chains have been extensively studied, it is imperative to recognize that human behavior changes may also amplify shocks to businesses and amenities that are connected via mobility and lifestyle patterns. In this talk, Dr. Yabe will present data-driven spatial network models to predict the cascades of shocks across cities, and further discuss applications to optimize civil infrastructure systems to achieve urban resilience. Dr. Yabe will also discuss his ongoing research on cross-city transfer learning approaches to prepare cities for unprecedented shocks.

Dr. Yafeng Yin

(UMich)

LLM-Agent-Based Simulation for Travel Demand Forecasting

(Recording) (Slides)

This talk explores the potential of leveraging LLM agents to advance activity-based modeling of travel demand. Traditional agent-based simulations rely on hardcoded condition-action rules for agents to schedule daily activities and make travel plans. These agents adapt and interact within dynamic environments but are constrained by predefined behavioral assumptions. Dr. Yin proposes replacing these agents with LLM agents that use large language models as reasoning engines. Unlike traditional agents, LLM agents derive decision rules from the parametric knowledge embedded in the LLM, eliminating the need for manual rule-coding. This paradigm shift could enhance activity-based microsimulation by relaxing behavioral assumptions, reducing data requirements, and offering greater flexibility to evaluate innovative plans and policies.

Dr. Xilei Zhao

(UF)

Enhancing Travel Behavior Analysis with Generative AI: Integrating Survey and Location Data

(Recording) (Slides)

Household travel survey data have long been essential for analyzing travel behavior. While valuable, these data face major limitations, including high data collection costs, small sample sizes, and insufficient spatial details about trips. Mobile device location data, on the other hand, have recently gained attention in transportation planning. They provide large-scale, granular travel trajectories at a low cost but lack critical traveler information, such as sociodemographics and preferences/attitudes. This talk introduces a novel Generative AI framework that integrates household travel survey data with mobile device location data. By combining their strengths, this approach addresses existing limitations and paves the way for next-generation transportation planning.

> > > Workshop Posters

List of posters