Amsterdam has the ambition to become accessible for all citizens. Unfortunately, the freedom to move without physical barriers is not the case for everyone yet. Some citizens with disabilities face physical and mental barriers in the urban infrastructure. The ’Amsterdam for All’ initiative is brought into life to address blind spots in urban accessibility with the help of emerging technologies such as artificial intelligence (AI).
In the first blog posts from the Amsterdam for All project, we have already addressed the scope and research directions. Within the current research phase, we focus on the walkability of the urban infrastructure. This is because, for people that experience physical disabilities, a sidewalk, crossing, or venue can be a barrier to their freedom of movement. Sidewalks can be narrow and are often covered with permanent and temporary obstacles. Our objective is to use data and AI to research and map the residual sidewalk space within Amsterdam. This is to measure and map the urban walkability of Amsterdam.
Altogether, within this article we will address previous research on this topic, already existing data, and several opportunities to address gaps in the data with AI.
Walkability research
The walkability is based on the combination of effective walking space and pedestrian demand [1]. Source: Amsterdam Intelligence
From 2016 till 2018, the Transport and Public Space department (Verkeer & Openbare Ruimte) of the city of Amsterdam has executed a pilot on researching the walkability of sidewalks. During this pilot, a universal methodology was designed to calculate a walkability score for sidewalks in a selected region of Amsterdam. The walkability index was based on two parameters shown in the figure above: 1. the effective walking space, 2. pedestrian demand. The effective walking space was determined with the ‘Basisregistratie Grootschalige Topografie’ (BGT), which is a national database containing geo-information on urban infrastructure design like roads, buildings, and trees. [2] In the figure below you can find the sidewalks visualized (in blue) with its permanent obstructions as trees and venue terraces. In the subsequent visualization the resulting walking space is shown.
Sidewalk with objects from available data sources [1]. Source: Amsterdam Intelligence
Available sidewalk space based on sidewalk width and its permanent obstacles. [1] Source: Amsterdam Intelligence
The pedestrian demand was estimated by neighborhood characteristics such as schools, enterprises, and public transport stations. Together, the space and demand of the sidewalk result in the overall walkability score. [1, 3] As you can see, the current walkability map does not cover the whole city. Therefore, the Amsterdam Intelligence team will continue this research and scale this work to provide the residual sidewalk width of the entire city.
Bike and Sidewalk network
Recently, OIS (Onderzoek, Informatie en Statistiek) of the city of Amsterdam has developed a graph network consisting of all bike and walking paths, called Loop- en fietsnetwerk (bike- and walk-network) Amsterdam. This graph has been developed for research and analysis purposes like determination of the proximity of services (parks, litter containers, or public transport), shortest distance, or route optimization. The nodes of this network furthermore contain information on the sidewalk width. However, it does not yet contain the objects on the sidewalk, which is something we would like to address and add to. The output could be added to Open Street Maps so that third parties could jump in. For instance to improve the routing for minorities which require enough sidewalk space. [4]
Available data sources
We need information on all objects that can be an obstruction to define walkability more accurately and provide a better representation of the width of the sidewalks. The objects on sidewalks can be either temporary or permanent. A parked bike on the sidewalk is for example a temporary hindrance but stairs can act as a permanent bottleneck. For obstacles without available data, AI can play an important role in detecting and registering on the 3D point cloud data of the city. Our articles of the point cloud project can be found here.
The most important dataset for this project is the ‘Basisregistratie Grootschalige Topografie’ (BGT), which is a national database containing geo-information on urban infrastructure design like roads, buildings, and trees. From this source, we can use the shapefiles of each sidewalk to determine the width. Furthermore, it contains the exact location of urban objects like streetlights, litter containers, and stairs. This information of permanent nature helps us more accurately define the residual sidewalk width. For some objects, the BGT is incomplete or lacks accuracy. Moreover, a few obstacles like the famous little ‘Amsterdammertjes’ poles are not registered in this dataset. Therefore, we would like to detect incomplete and missing data. Currently, the AI team is researching the feasibility of detecting objects by using computer vision on the 3D point cloud data. Ideally, these detections will be added to the BGT. The early results look promising, and we expect to detect unregistered objects, improve the completeness of the BGT, and therefore urban infrastructure information.
Bike crowdedness prediction
Bike crowdedness per sidewalk [1]. Source: Amsterdam Intelligence
Temporary objects on sidewalks are not saved in data. Therefore, we aim to research the ability to predict the presence or incidence of obstacles like bikes in the street. Yearly, The Transport and Public Space department (Verkeer & Openbare Ruimte) executes a bike count including registering the bike rack capacity and its occupation in the city on street level. This data in combination with neighborhood characteristics and panoramic (2D) or point cloud (3D) images will be used to predict the bike occurrence for time and place in the city. This would make it possible to provide information on the crowdedness of bikes per sidewalk.
Conclusion
All in all, there is plenty of data available to approximate the residual sidewalk width and space. The challenge will be to validate the accuracy of the data and improve the quality. Further research and implementation of machine learning on 2D and 3D images will help to register the data gaps and improve the accuracy of the data sources. With this approach, the Amsterdam for All initiative aims to inform city officials on urban accessibility levels. Moreover, we thrive to add this extra layer of information to existing route planning services so this information reaches third parties. Then, more personalized routes can be offered to our citizens.
In addition, this work can serve the city on a broader level than accessibility only because it will improve the availability of urban infrastructure data. This can be of great use for urban planning and making the city a better livable city for all. Eventually, we admire sharing and scaling this methodology to other Dutch municipalities.
References
[1] Walkability in Amsterdam Presentation, Manchester, Wednesday November 21th, 2018. Eric de Kievit | Senior Advisor Transport & Traffic Research | Mobility & Public Space
[2] BGT website, https://www.pdok.nl/introductie/-/article/basisregistratie-grootschalige-topografie-bgt-
[3] Walkability maps, prototype. Website Spacetraces: https://www.spacetraces.com/walkability-prototype
[4] Loop- en Fietsnetwerk Amsterdam, https://data.amsterdam.nl/datasets/7hGzsRXqWSGqHw/loop-en-fietsnetwerk-amsterdam/
Dit artikel is afkomstig van de website van Amsterdam Intelligence