A Crowdsourcing Platform for the Construction of Accessibility Maps
Carlos Cardonha, Diego Gallo, Priscilla Avegliano, Ricardo Herrmann, Fernando Koch, Sergio Borger · 2013 · Proceedings of the 10th International Cross-Disciplinary Conference on Web Accessibility (W4A) · doi:10.1145/2461121.2461129
Summary
This highly cited paper (69 citations) from IBM Research Brazil presents a comprehensive crowdsourcing platform for constructing outdoor accessibility maps through two complementary data collection approaches. The first, IBM Citizen Sensing, is an active reporting tool where users manually submit geolocated reports about accessibility barriers (with photos, categorised attributes, and text/voice comments). The second, IBM Breadcrumb, is a passive data collection tool that automatically captures periodic measurements of geolocation, acceleration, orientation, and device ID from smartphone sensors without requiring user interaction — essentially turning pedestrians into walking accessibility sensors. The platform integrates data from both tools with analytics modules for statistical analysis, simulation, prediction, and optimisation to generate accessibility maps that augment standard mapping services like Google Maps and OpenStreetMap with an accessibility information layer. The authors argue that the relevance of an accessibility problem is proportional to its impact on citizens' lives, making crowdsourced data from affected people the most appropriate way to catalogue and prioritise urban accessibility issues.
Key findings
A field experiment in São Paulo demonstrated the platform's potential. A person walked a route with Breadcrumb collecting GPS and altitude data every 10 seconds, while accessibility issues along the same route were reported using Citizen Sensing. Analysis of the Breadcrumb data — specifically walking speed estimated from GPS coordinates using simple moving averages with outlier capping at 6 km/h — revealed that speed decreased at four points along the route. Three of these (A, B, C) corresponded to Citizen Sensing reports of accessibility barriers: no sidewalk at all, a light post obstructing the sidewalk, and steps in the sidewalk. The fourth (X) was a road crossing where the person stopped, demonstrating a key insight: individual deviations (like stopping at a crossing) become insignificant with data from many people, while genuine accessibility obstacles would slow nearly everyone taking that path. The authors identified an important confound: altitude changes also affect speed, so slope must be accounted for before inferring accessibility problems from speed reductions. The paper outlines future analytics directions including correlation analysis (detecting systematic report misclassification), sentiment analysis (extracting emotion from text/voice reports to gauge severity), prioritisation using Analytic Hierarchy Process, and accessible route planning that minimises expected accessibility barriers.
Relevance
This is the most cited paper (69 citations) among the recent batch and represents a significant advance in accessibility mapping methodology. While the companion demo paper (IBM Sidewalks, also in the database) focused on the active reporting app, this paper's key contribution is the Breadcrumb passive sensing concept: using smartphone accelerometers and GPS to automatically detect accessibility barriers through changes in pedestrian walking patterns, without requiring users to actively report anything. This passive approach addresses the fundamental scalability limitation of active reporting — most people won't bother to stop and file a report, but they'll carry a phone that passively logs their movement. For smart city initiatives, the paper provides a practical framework for combining active citizen reports with passive sensor data and analytics to create dynamic accessibility maps. The vision of accessible route planning — finding paths that minimise expected accessibility barriers — anticipates features now being developed in mainstream navigation apps. The concept of dynamic accessibility maps that account for temporary barriers (construction, flooding) is particularly forward-looking.
Tags: crowdsourcing · accessibility mapping · physical accessibility · urban accessibility · citizen sensing · smart cities · mobile accessibility · data visualization · sensor data