Benjamin Tannert

Marvin Lange, Reuben Kirkham, Benjamin Tannert

There has been a considerable amount of research aimed at automating the documentation of accessibility in the built environment. Yet so far, there has been no fully automatic system that has been shown to reliably document surface quality barriers in the built environment in real-time. This is a mixed problem of HCI and applied machine learning, requiring the careful use of applied machine learning to address the real-world concern of practical documentation. To address this challenge, we offer a framework for designing applied machine learning approaches aimed at documenting the (in)accessibility of the built environment. This framework is designed to take into account the real-world picture, recognizing that the design of any accessibility documentation system has to take into account a range of factors that are not usually considered in machine learning research. We then apply this framework in a case study, illustrating an approach which can obtain a f-ratio of 0.952 in the best-case scenario.

Reuben Kirkham, Benjamin Tannert

Navigating from 'A to B' remains a serious problem for many people with mobility impairments, due to the need to avoid accessibility barriers. Yet there is currently no effective routing tool that is regularly used by people with disabilities in order to effectively avoid accessibility barriers in the built environment. To explore what is required to produce an effective routing tool, we have conducted Monte-Carlo simulations, simulating over 460 million journeys. This work illustrates the need to focus on barrier minimization, instead of barrier avoidance, due to the limitations of what can be achieved by any accessibility documentation tool. We also make a substantial contribution to the concern of meaningful performance metrics for activity recognition, illustrating how simulations can operate as useful real-world performance metrics for information sources utilized by navigation systems.