Rad Roads using OSMnx

Credit: Yuwen Chang, YuChen, and Cheng Hou
An ARGO civic data marketplace project.

Rad Roads in Queens & Bronx, New York.

RadRoads in Bronx, NY

The top-5 longest walkable roads:

1. Bruckner Boulevard @ 17.5 miles (both ways)
2. Grand Concourse @ 14.83 miles
3. East Tremont Avenue @ 8 miles
4. White Plains Road @ 7.16 miles
5. Boston Road @ 6.5 miles

Amongst the shortest roads in the Bronx are Woodmansten Place

Amongst the straightest roads in the Bronx is Maran Place

Rad Roads in Queens, NY

The top-5 longest walkable roads:

1. Queens Boulevard @ ~25 miles
2. Union Turnpike @ 19 miles
3. Francis Lewis Boulevard @ 17.5 miles
4. Northern Boulevard @ 14.5 miles
5. Horace Harding Expressway @ 14 milles.

Amongst the shortest roads in Queens are Fowler Avenue & Gravett Road

Amongst the most straightest roads in Queens is Westmoreland Place.

These roads were identified using RadRoads, a project that aims to creatively explore road networks in a city, programmatically.

Road networks define the rhythm of a how a city moves. From walkability, to access to jobs, our cities’ roads are the proverbial needles that weave our civic fabric.

The recent fervor in automating road transportation has largely left out the role of the design and maintenance of roads. The promises of vehicle autonomy span from the saving of lives to an increase in productivity yet the reality of vehicle autonomy lies in favor to the more privileged and lacks an organizing framework to serve public needs, at scale.

Consider that navigating a complex street grid is computationally more expensive than one that is dead straight. Even with the availability of high fidelity maps and advanced sensing instruments, an autonomous vehicle relies on the presence of road markings to move safely on any road.

Curved roads can be more complex than straight ones.

What if:

• These markings don’t exist on roads that serve low-income neighborhoods?
• Roads in low-income areas turn out to be less straighter and therefore more complex to navigate?
• A small-sized city wants to pilot autonomous vehicles on its streets? Where should it begin? Where can the benefits of vehicle autonomy most pronounced?

There are no clear answers to these questions and answering them in a manner that uses data and evidence is harder still.

Rad Roads is a step towards answering some of these questions with the goal of empowering cities to, using open data and open source technology, interrogate their street grid for various purposes.

Using data from Open Street Maps and OSMnx, RadRoads creates a set of consistent parameters so that different cities’ roads can be compared for application to, amongst other things, vehicle autonomy. RadRoads are part of a suite of Street Network analysis tools.

• ARGO’s Geo Street Talk demonstrated how we can use open data to enable a conversational interface between a person and an autonomous vehicle.
• SQUID Routing is an attempt to democratize routing a vehicle across each and every street in a city.

These efforts form a larger vision of building the digital infrastructure around a public service autonomous vehicle.

With RadRoads, we begin with answering basic questions about a city’s street grid:

• What is the longest road in a city?
• What is the shortest?
• How about the straightest or the curviest?

OSM & OSMnx primer

Open Street Maps (OSM) is a collaborative and freely editable mapping platform launched in 2004. Think of it as the Wikipedia equivalent for online mapping. Learn OSM is a great way to get started on OSM. It

OSMnx, developed by Geoff Boeing, is a combination of network and geospatial analysis that combine python packages like networkX, geopandas, matplotlib to analyze OSM data. The data is extracted from OpenStreetMap through the Nominatim API and converted into a network graph. It is very well documented with tons of example notebooks and easy to use.

Here’s a graph of driveways in a typical Brooklyn neighborhood.

A typical Brooklyn neighborhood as viewed through OSMnx

RadRoads Usage

RadRoads accepts a place name and network type as inputs:

• Place name follows the Nominatim naming syntax. For example: Los Angeles, California does not work but Los Angeles, CA works.
• Network type can include: drive - drivable public streets sans service roads, drive_service - drivable public streets + service roads, walk - all streets and paths that pedestrians can use , bike - all streets and paths that cyclists can use, all - all non-private OSM streets and paths, all_private - all OSM streets and paths, including private-access ones.

Depending on the areas you are trying to analyze, downloading can take some time. Downloading the road network from Manhattan, for instance, would take several minutes to complete. However, you may save the graph as shapefiles for offline use. The network graph also contains basic statistics, including both network statistics (those originally in NetworkX, such as average betweenness centrality) and some road statistics specifically added in the package (such as average street length).

RadRoads metrics

Length — The longest and shortest roads in a place.

Here, we combine all the segments under the same name and calculate the total length in meters.

In Manhattan for instance, the longest road, after combining all the segments, is Broadway with a total length of 33,808.79 meters ~ 21 miles.

RadRoads applied on Manhattan, NY

Straightness / Curviness

We used the concept of sinuosity (often used to measure the curviness of a river) to measure curvature and straightness of roads.

Wile this may work well for rivers, roads can be a different. Take a look at the following example, both routes (red lines) connecting A and B have similar lengths and, thus, roughly the same sinuosity. But which one do you think is “curvier”?

A road that detours more is less efficient in terms of land use, fuel consumption, in-vehicle time, etc.

Sinuosity of Roads. The one on the left is more sinewy and perhaps more complicated to navigate in an autonomous vehicle.

Kinks

Since RadRoads uses data from Open Street Maps, a community sourced open source mapping platform, there can be a few kinks. Sometimes the shortest or straightest road can end up being a small lane or a private driveway because they were marked incorrectly as a residential street. Improving the underlying Open Street map data would improve the quality of RadRoad results.

Perhaps RadRoads can be repurposed as a way to validate the quality of Open Street Map Data?

Moving Forward

We hope to apply RadRoads for the following use-cases:

• Route discovery and optimization.
• Site selection based on location accessibility, vulnerability and resiliency.

We also need to think about other features and statistics that can be applied to city streets to derive meaningful street insights. This depends on the questions that we ask and the stakeholders involved. A possibility is to look at the efficiency of street network in its entirety. This may involve combining street network data with population and business data.

Combining this analysis with the Census’ Longitudinal Employer-Household Dynamics could provide tremendous insight into deciding where to deploy autonomous public transit or other public services.

We invite you to help explore the world’s RadRoads with this. You can start by creating an issue on RadRoad’s Github repo.

argo-marketplace/RadRoads