Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deep Activity Model: A Generative Approach for Human Mobility Pattern Synthesis

Xishun Liao, Qinhua Jiang, Brian Yueshuai He, Yifan Liu, Chenchen Kuai, Jiaqi Ma

TL;DR

A novel generative deep learning approach for human mobility modeling and synthesis that incorporates both activity patterns and location trajectories using open-source data that can be fine-tuned with local data, allowing it to adapt to and accurately represent mobility patterns across diverse regions.

Abstract

Human mobility plays a crucial role in transportation, urban planning, and public health. Advances in deep learning and the availability of diverse mobility data have transformed mobility modeling. However, existing deep learning models often focus on spatio-temporal patterns and struggle to capture the semantic interdependencies among activities, while also being limited by specific data sources. These challenges reduce their realism and adaptability. Traditional activity-based models (ABMs) face issues as well, relying on rigid assumptions and requiring extensive data, making them costly and difficult to adapt to new regions, especially those with limited conventional travel data. To address these limitations, we develop a novel generative deep learning approach for human mobility modeling and synthesis that incorporates both activity patterns and location trajectories using open-source data. The model can be fine-tuned with local data, allowing it to adapt to and accurately represent mobility patterns across diverse regions. The model is evaluated on a nationwide dataset of the United States, where it demonstrates superior performance in generating activity-location chains that closely follow ground truth distributions. Further tests using state- or city-specific datasets from California, Washington, and Mexico City confirm its transferability. This innovative approach offers substantial potential to advance mobility modeling research, particularly in generating synthetic human mobility data. This can provide urban planners and policymakers with enhanced tools for simulating mobility in diverse regions and better informing decisions related to transportation, urban development, and public health.

Deep Activity Model: A Generative Approach for Human Mobility Pattern Synthesis

TL;DR

A novel generative deep learning approach for human mobility modeling and synthesis that incorporates both activity patterns and location trajectories using open-source data that can be fine-tuned with local data, allowing it to adapt to and accurately represent mobility patterns across diverse regions.

Abstract

Human mobility plays a crucial role in transportation, urban planning, and public health. Advances in deep learning and the availability of diverse mobility data have transformed mobility modeling. However, existing deep learning models often focus on spatio-temporal patterns and struggle to capture the semantic interdependencies among activities, while also being limited by specific data sources. These challenges reduce their realism and adaptability. Traditional activity-based models (ABMs) face issues as well, relying on rigid assumptions and requiring extensive data, making them costly and difficult to adapt to new regions, especially those with limited conventional travel data. To address these limitations, we develop a novel generative deep learning approach for human mobility modeling and synthesis that incorporates both activity patterns and location trajectories using open-source data. The model can be fine-tuned with local data, allowing it to adapt to and accurately represent mobility patterns across diverse regions. The model is evaluated on a nationwide dataset of the United States, where it demonstrates superior performance in generating activity-location chains that closely follow ground truth distributions. Further tests using state- or city-specific datasets from California, Washington, and Mexico City confirm its transferability. This innovative approach offers substantial potential to advance mobility modeling research, particularly in generating synthetic human mobility data. This can provide urban planners and policymakers with enhanced tools for simulating mobility in diverse regions and better informing decisions related to transportation, urban development, and public health.
Paper Structure (29 sections, 4 equations, 15 figures, 6 tables, 2 algorithms)

This paper contains 29 sections, 4 equations, 15 figures, 6 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Motivation
  3. Objective and Study Scope
  4. Contributions
  5. Related work
  6. Problem Formulation
  7. Dataset
  8. Household Travel Survey
  9. SCAG ABM Data
  10. Data Preparation
  11. Methodology
  12. Model Architectures
  13. Loss functions
  14. Data Balancing
  15. Model Transfer
  16. ...and 14 more sections

Figures (15)

  • Figure 1: Model human mobility pattern using HTS data. (a) HTS data includes information about each household member's socio-demographics, household characteristics, and daily non-commercial travel across all modes, including details about travelers, their households, and vehicles mcguckin2018summary. (b) Typical weekday and weekend activity chains in HTS.
  • Figure 2: Workflow of activity chain generation. Given the synthetic socio-demographic information and household characteristics of each agent, the model auto-regressively synthesizes the agent's activity chain and the location of each activity.
  • Figure 3: Deep Activity model architecture. (a) Input data construction. (b) Transformer-based network architecture with well-designed data injection.
  • Figure 4: Distributions of activity distances and angular deviations across sub-regions in LA
  • Figure 5: Detailed analysis and comparison of activity generation on (a)(b) temporal dynamics, (c) activity chain length, (d) activity duration, and (e) activity type distribution.
  • ...and 10 more figures