Inclusive Accessibility – A deep learning-based space-time framework integrating person-based hard and soft constraints

Keywords: accessibility, mobility, equity, GIScience, spatio-temporal analysis, data-science, deep learning, artificial intelligence
Session Type: Virtual Paper Abstract
Day: Tuesday
Session Start / End Time: 3/1/2022 08:00 AM (Eastern Time (US & Canada)) - 3/1/2022 09:20 AM (Eastern Time (US & Canada))
Room: Virtual 35

Abstract

Individuals evaluate multimodal accessibility based on their heterogeneous perceptions and everyday mobility experiences. This perceived accessibility of individuals generally differs from their physical accessibility, delimited using explicit and hard space-time constraints. This study refers social and built environment factors (e.g., personalized network and route preferences, lived experiences, and perception toward safety and built environment) as soft constraints since the influence of these variables on individual-level accessibility are implicit, perception-based, and rely on the decision-making strategies of individuals. This study investigates the impacts of soft constraints on the perceived accessibility of individuals using off-the-shelf deep learning techniques. The study will collect smartphone app-based survey data on individual-level mobility needs and travel experiences and fuse the survey data with data collected from crowdsourced mapping and mobility data platforms. We will model network-based spatio-temporal variability of soft accessibility constraints using deep algorithms such as multinomial logistic model and neural network and evaluate their performances. The estimated weights will indicate the extent to which different social and built environment characteristics affect the perceived accessibility constraints and their variability across individuals. Later, the model predictions will help us develop an inclusive accessibility measure integrating both hard and soft constraints to examine the difference between physical and perceived accessibility of individuals. This GeoAI-based convergence of hard and soft constraints will develop more accurate and inclusive measures of accessibility, strengthen equity-oriented urban transportation science and planning, and promote need-specific transportation planning, social equity, well-being, and health.