AI for measuring the quality of walking environments?

Recently, Kang and colleagues tested neural networks to assess and predict walkability . They:

• Automatically gathered street view images
• Collected people’s preferences (people compared pairs of images selecting the image that was “good for walking“; there were 52 participants and disability is not mentioned, so they probably were non-disabled)
• Built a neural network model that could predict people’s preferences based on the similarities and differences between images
• Assessed the “physical walkability” but found that there is no consistency in the literature in how to measure it and noted difficulties. This happens to be the area that most interests me!
• Compared physical and perceived walkability with the issues related to the physical walkability.

How to assess the quality of walking environments?

Kang and colleagues used four categories “safety, convenience, comfort, and accessibility, which were the most frequent and considered reasonable“ – note that these do not include feasibility or pleasantness, respectively at the bottom and top of perceived pedestrian needs, according to the Social Model of Walkability.

For reach category they chose indicators, reported below, again finding that the literature is inconsistent. Note that none of the indicators include traffic or ease of crossing (which should be part of feasibility, accessibility, safety and comfort) or physical obstacles such as sidewalk obstructions, insufficient width, or uneven surfaces.

The authors suggest that “The difference between the scores of physical and perceived walkability provided implications for improving the walking environment. In other words, considering the scores of physical and perceived walkability, the areas with high scores of physical walkability but relatively low scores of perceived walkability were judged to be where the walking environment should be improved in priority.”

They however note (p. 20):

Although we successfully predicted the perceived walkability score using a deep learning model based on the pairwise comparison dataset, we could not fully understand the factors that influence perceived walkability scores. Investigating the underlying factors that affect perceived walkability is crucial, not only in the domain of walkability research but also in other fields that employ deep learning techniques.

I can see so much potential in this type of approach but it would require assessing physical walkability in a more accurate way. And that's precisely the aim of the workshop series we are organising together with (by anti-alphabetical order) the Walk21 Foundation , Szening Ooi , Louise Reardon , Karen Seaman, Josephine Roper, Carlos Canas, Bronwen Thornton , Belen Iturralde , and Alan Meharry .

We already ran the first workshop, on the theoretical model underpinning the relationships between walking environments and behaviours, and are going to run the next one soon – on ways to measure physical walkability, and in particular the quality of street environments.

If you are interested in participating, don’t hesitate to fill the quick registration form here , and you will receive an invite in the next days!

#walkability #walking #streets #streetsforpeople #citiesforpeople #retrofit #urbandesign #StepSworkshops