Throughout the 20th century, most modern cities have followed a development with a strong focus on motoring. This is the introductory statement in a recently published article: “Data-driven strategies for optimal bicycle network growth”, written by the international research team Luis Guillermo Natera Orozco, Federico Battiston and Gerardo Iniquez, led by Michael Szell from the IT University in Copenhagen . In research, cities are represented as networks. Using network algorithms, our research team spotted how small investments in bike paths can greatly improve the transportation experience for cyclists.
There is no doubt that motoring is leaving its mark on urban areas in the form of air pollution, traffic jams during rush hour and negative impacts on the environment. Many of these problems can be largely eliminated if you replace cars with bicycles and public transport. Expanding the infrastructure for cycling is thus a healthy investment for urban areas.
Expansion of the bicycle network
What is missing in relation to an improvement of the bicycle infrastructure are the links that must connect one part of the city with another and ensure that one can transport oneself easily, quickly and safely through the city – e.g. via dedicated bike lanes.
Illustration of how to convert a map of a city to a mathematical object called a network or graph. If we draw blue lines (called joints) between intersections – here represented as black dots (so-called nodes) – we get a network representation of the city.
Illustration: From city map to network
The problem is that the cycle paths in many cities are spread out over the city in small, isolated ‘groups’. Let us imagine that you will take the bike to ride out to shop because there are good and safe bike paths in your local area. But you can not take the bike to the district right next to where the mall is located. From a cycling perspective, these neighborhoods and areas can be seen as tiny little networks that should be connected to one large network that covered the entire city. We call these small networks ‘components’, and by connecting them to a single cycle path (a joint), they come together into a large component network – which suddenly means that you can now take a trip to the mall.
So questions are: Why hasn’t your city just established bike paths between neighborhoods? The answer is that it is primarily because politicians are sluggish in their upbringing. But it is also because establishing cycle paths is a task that requires careful planning.
The decision on where to have new bike paths
With a good cycling infrastructure, you can cycle wherever you want in a safe way. If we speak ‘network language’, it means that the network must be well connected. And in addition, it is important that the route is fairly direct: Why ride 10 km by bike if you can get to the same destination by car and only cover 5 km?
Here, the word ‘direct’ describes the degree to which it is easy to take the bike instead of the car. The more direct and connected a bicycle network can become, the better. But unfortunately, it is not free to establish new bicycle infrastructure. So we have been looking for methods that can increase the degree of direct connections with minimal investment.
To find optimal locations for new bike paths, we experimented with several different growth algorithms.
The illustration shows a network that could represent a city. To begin with, it consists of four separate components. Where each step the algorithm finds the largest component and connects it with the nearest neighbor.
Algorithms that predict growth are used to expand networks. In our research, each algorithm determines which two intersections should be connected to a bike path. After simulating 5 km of hypothetically possible, new cycle paths, our research team compared how much the different algorithms had improved the degree of direct connections.
By simulating the same process for 15 cities in different countries, our research, based on the different approaches, identified the theoretically best strategy for planning new cycle paths: namely to connect a city’s largest, already existing network components with the nearest other component. The two components thus become one combined, even larger component, and then this process is repeated. One can see it as building a lot of short bike paths that act as bridges between well-functioning islands of unified bike paths.
Results vary from city to city
However, the 15 cities do not experience the same improvements, as they have very different starting points. Cities that already have a well-developed cycling infrastructure will get more value for money with additional investment. The cities that get the most out of small investments are the cities that already have well-established cycling infrastructure, such as. Copenhagen, Amsterdam and Portland. These cities have only small gaps in their networks, and by filling these, the work that has already been done can unleash its full potential.
People want more than just efficiency
Infrastructure planning cannot be based solely on dots and lines in a computer simulation. In urban planning, one must also take into account human behavior and the very real factors of reality. For example, people may not bother to drive miles uphill, or it may be that they think green areas along the bike paths are nice. Therefore, this research is only the first step towards a more quantitative coverage of how to develop bicycle infrastructure, as it only examines the problems of urban planning from a technical point of view, and (yet) does not include human factors – as f. ex. what residents and cyclists really want.
Through our research work, we have scaled the study up to include 60 cities and further expanded our focus on connectedness and the individual links to also include the growth of the overall network in most of the individual city. In addition, we use scientific methods about networks to identify and classify gaps in the cities’ bicycle networks – here we have started with Copenhagen. And to also include the behavioral factors, we explore methods in image recognition and machine learning, so you get some automated methods to analyze how cyclists actually prefer to move around the city.
Here you can read the research article in the Royal Society Open Science .