Walkability: Agent Based Models

Introduction

In this tutorial, we will introduce a walking demonstrator tool which will help you understand walkability and public transport using Ped-Catch. By setting simple parameters such as maximum walking time, walk speed and crossing wait time, it generates a piece of animation showing how the intelligent agents walk around the road network (Check out this link for more details)

Using the Demonstrator

This demonstrator is available here

The link above should bring you to the following page

[Click to Enlarge]

[Click to Enlarge]

 

The web page contains two main parts: a control panel and a map view. In the control panel, we can set up three parameters which control the agents’  navigation behaviour: maximum walking time, walk speed and crossing wait time. The simulation results will be rendered on the map view, which contains a layer controller. By default, it shows the road network layer, and we can turn on the destination layer (red dots).

To start a simulation, we need pick up an origin on the map. To do this, just click anywhere close to the road network. Tips: the origin point doesn’t have to be placed on the road, walkability algorithm will automatically create a path connecting the origin to the road network. But if the origin is too distant away (say 300 meters), the algorithm will fail to do so and you will see an error message “Generated output is empty” above the map view.

[Click to Enlarge]

[Click to Enlarge]

 

Let’s choose Melton train station as our origin point:

[Click to Enlarge]

[Click to Enlarge]

 

Then click the green “Simulate” button to run the algorithm. The computing time largely depends on the “Maximum Walking Time”. The longer walking time, the longer computing time is required.

When computation is done, you will see a big red circle on the map. The disabled “Play” and “Pause” buttons are now active.

Click the “Play” button to see the walkability animation. You will see circles spreading over the road network in three colors: red (0<= walk time <4 minutes), orange (4 minutes <= walk time < 8 minutes) and yellow (8 minutes <= walk time).  Here are some example outputs:

maximum walking time = 20 min; walk speed = 1.33 m/s; crossing wait time = 30 sec
[Click to Enlarge]

[Click to Enlarge]

[Click to Enlarge]

[Click to Enlarge]

Once a simulation is finished, you can choose another origin point on the map and perform a new simulation and then play the animation. You can turn the results of multiple runs on and off in the layer controller.

NOTICE: If multiple outputs are shown on the map, only the last animation can be replayed.

 

walk-ui6

[Click to Enlarge]

More Information

Understanding Walkability and Public Transport Using Ped-Catch

People

Co-led by Dr Hannah Badland, McCaughey Centre, VicHealth Centre for Promotion of Mental Health and Community Wellbeing, and Dr Marcus White, Faculty of Architecture, Building, and Planning.
Victorian Government Champions: Jim Betts, Secretary, Department of Transport; and Christine Kilmartin, Manager, Sustainability Analysis, Department of Planning and Community Development.
Research Champion: Professor Billie-Giles Corti, Director, McCaughey Centre, The University of Melbourne.

About

This project will develop a walkability index which will be calculated and applied to census collection districts surrounding public transport nodes in the North West Metropolitan Region. This agent-based modelling tool has the capacity to be not only a powerful urban design tool that builds on existing walkability measures, but also an influential advocacy tool. The purpose of the tool was to yield a more accurate understanding of how neighbourhood walkability is associated with access and permeability, and to develop an interactive on-line tool for researchers and planners to modify neighbourhood walkability to enhance access to features of interest. As such, this work will provide not only innovative tools to investigate how neighbourhood walkability is related to amenity access, but enables different planning scenarios to be tested prior to developing new or retrofitting older areas. It is anticipated that planners will apply these tools to diverse areas in Melbourne’s North West Metropolitan region and beyond, prior to building infrastructure or when seeking to modify existing sites.

A strength of this tool is its spatial data flexibility; that is, different users have the ability to upload different data sources at different scales. In order to do this, the tool has been developed with a spatial data hierarchy in mind. Fine-grained data are optimal, but inputs extend too coarser-scale (e.g., SA2-level) and open-access (e.g., Walk Score) data sources. In this way, a variety of different end users are able to utilise the tool either using their own data, or those supplied through the AURIN portal or other open-access sources. Currently, standard datasets used for the tool include the road network, features of interest (e.g., schools, public transport nodes), and traffic lights. Depending on end user access to other spatial data, additional spatial layers can also be incorporated into the tool. These include footpaths, traffic volume, and topography. Including such additional spatial data layers enhances the accuracy of the tool. There is also scope to add in more subjective features of the environment, such as shade, crime and incivilities, and aesthetics.

User-specified functionality

This was conceptualised as being a flexible tool that allows users to test different scenarios based on features of interest (e.g., public access nodes, school locations), street connectivity, and population of interest (e.g., vulnerable populations, such as children or older adults). In order to achieve this, a series of user-specified functionalities were designed into the interface.
These included sliding bars to manipulate the: maximum walking time (up to 20 minutes), maximum walking speed (up to 2 m.s-1), and intersection wait time (up to 60 seconds). Vector editing tools were also provided in the interface, allowing the user to: add or remove street networks to modify street connectivity; and manipulate the agents starting point to reflect potential features of interest (e.g., public transport egress, location of a school). These attributes are theoretically linked to walking behaviours. For example, walking speeds vary greatly with different ages and levels of mobility, and having the ability to alter pedestrian speeds was important for investigating potential for different nodes of interest (e.g., primary schools, senior citizen organisations). It was hypothesised such destinations would have smaller catchments than others due to the different expected walking speeds.

Vector editing functionality

As per limitations of earlier walkability tools and feedback from stakeholders, there was a recognised need to include vector-editing functionality. This would enable the user to modify the street network and test different scenarios prior to retrofitting an environment. This was regarded by the stakeholder working group as being a valuable extension to the tool, and was created by snapping vectors to the existing street networks. Multiple vectors can be added or removed within a given scenario. An example of this is shown in Figure 1, where the blue line in top image has been added by the user; the image below shows the agents travelling the new connection.

[Click to Enlarge]

[Click to Enlarge]

[Click to Enlarge]

[Click to Enlarge]

Figure 1. Agent-based model screen-shots showing vector editing features

Broader model considerations

As well as providing user and vector editing functionalities, broader considerations for model development included having to: operate within an open-source environment, and within the AURIN portal requirements and architecture; be flexible enough to include a range and hierarchy of spatial data and scales provided by the end user; function in hardware with lower computational power; and provide an interface that was easy to navigate. In order to do this, the tool was developed using basic agents with a limited level of artificial intelligence. Agents left from a user specified node and travelled to a randomly distributed ‘cookie-crumb’ snapped to the street network within the parameters set by the user (e.g., walking speed, time, intersection wait time) (see Figure 2). Having a limited artificial intelligence ensures different ‘what if’ scenarios can be rapidly tested; the stakeholder group regarded this as being an important feature.

[Click to Enlarge]

[Click to Enlarge]

[Click to Enlarge]

[Click to Enlarge]

[Click to Enlarge]

[Click to Enlarge]

[Click to Enlarge]

[Click to Enlarge]

Figure 2. Agent-based model screen-shots

Outputs include: a visual, graded representation of agents throughout the network; area coverage comparison between the agent-based model and circular catchment expressed as a ratio; and mean data on the number of intersections crossed. All of these variables are recognised as being important attributes of walkability. As well as a map (e.g., Figure 2), these output data are generated as a .csv file after running each model, thereby allowing comparisons of walkability to be made across different hypothetical built environments. Together these outputs enable the user to understand how walkability is influenced by built environmental and behavioural modifications, or test specific environments to reflect the population of interest (e.g., children, older adults).

Licensing

Because of the open-source nature of the current data, theoretically the agent-based model can be used across most built environments internationally. A online simulation of the agent-based model can be found here: http://115.146.93.38:9999/agent-walkability/agent-model.html#about, with another example here: http://115.146.87.16:9999/agent-walkability-bb/agent-model.html

The tool is designed to be overlaid with the walkability index, which is housed in the AURIN portal and is created by the Place, Health, and Liveability Program, University of Melbourne.

This project has been supported by the University of Melbourne Centre for Spatial Data Infrastructure and Land Administration (CSDILA), the Victorian Government Office of the Valuer General, the Australian National Data Service (ANDS) and the Australian Urban Research Infrastructure Network (AURIN) through the National Collaborative Research Infrastructure Strategy Program and the Education Investment Fund (EIF) Super Science Initiative.