How can cities become more sustainable and less car dependent while still remaining accessible to all?
This challenge drove the development of the D4AMS online dashboard. For this project, we are interested in the role of specific mobility interventions in realising this challenge.
We look at two specific types of mobility interventions:
By playing with these kind of interventions, we simulate alternative mobility scenarios to generate specific policy guidelines.
This dashboard was developed using Agent Based Modelling. Specifically, we used the MATSim model (Multi-Agent Transport Simulation). This is a computational modelling technique that, as the name suggests, is based on multiple agents, or persons. These agents are autonomous elements, which live, move and interact in a city. They can help us understand certain travel behaviour. In this research, we are interested in studying collective travel behaviour.
Two aspects are present in the model:
Plans or goals
Each agent has a day plan consisting of activities, for example going to work, grocery shopping, eating out, etc. These activities need to be carried out at a specific location and time. Reaching these activities is done through travel, with a transportation mode, using a certain route. Plans can be based on real data, such as travel diaries, or on a 'synthetic population', which is an imaginary population with characteristics derived from census data.
Rules
In this model, agents are subject to rules about what they are allowed to do and what not to do, given their personal plans. Which modes are allowed to be used, which streets are accessible for which mode and how fast are people allowed to drive? These rules are based on the existing street network (where can someone drive?) and on current policies (are there sharing facilities available?).
In this model, agents try to perform their daily plans. When doing so, each agents gets a utility score based on 'how good' or 'how useful' their daily plan was carried out. Basically, positive utility is gained through performing activities and negative utility is gained by travelling. The longer an agents travels, the lower the utility score will be. For example, using a car is seen as more convenient then using public transport, resulting in a higher initial score for car travel.
The model runs until each agents has an optimal score. Every time a new run is carried out, the model slightly adopts the agents' plans (for example trying out a different transportation mode, leaving a little bit earlier or using a different route), to see if the scoring can be increased. For example, if an agents gets stuck in traffic with a car during the first simulation, that agent will try a different route in the second simulation. This, in turn, will have an impact on other agents also using that route. In the third simulation, that agents might choose for an alternative mode, a bicycle for instance. This will diminish the traffic jam on the route used. In this way, agents change there behavior collectively.
Now, this setting allows us to experiment with possible changes in the model: we can change the network (the streets) or we can change the rules or opportunities (e.g. which modes can be used, speed limits, etc.). This is exactly what we do for creating alternative mobility scenarios: we implement street closures (adapting the network) and we install mobility sharing hubs (adapting the rules). These changes will alter the way how agents are allowed to travel and, thus, result in different traffic outcomes.
Through this model, also other types of changes could be hypothetically simulated, such as adding a cycling bridge, implementing cycling or 30km/h zones, extending public transportation services to less connected areas, etc.
The maps show the change in traffic flows for each scenario. The maps are divided in hexagons. This is done because the original input data represents only a sample of the whole city population. Therefore it would be scientifically incorrect to make statements on the street level and would it be better to look at a higher scale, in this case the neighbourhood level, represented by a hexagon. By dividing the city into hexagons, it is possible to make statements about more parts of the city, albeit slightly more generalised. As noticeable, no specific numbers or percentage of traffic flow are given. This is because we are examining trends of hypothetical scenarios, rather then absolute outcomes. Based on all the assumptions and limited sample size used in this traffic model, exact percentages and numbers would overestimate the abilities of our dashboard and prove to be incorrect.
Simulating scenarios using this model is valuable because it requires no real world testbed to evaluate certain changes applied. Instead, changes can be simulated, using real-world data.
Of course, this comes with some assumptions made, as a model is an abstracted representation of the reality.
One of the main assumptions is that people behave and travel as 'homo economicus'. This concept sees humans as a species who wants to maximize its personal utility, while being fully rational and having full knowledge of all possibilities. The Homo Economicus is an antisocial species that acts only economically. This is an irrational representation of reality that leads to a deterministic view of humanity, because people also act emotionally or randomly and full knowledge is impossible. For example, some people will always take the same route, even though they know it is not the shortest, but just because they like it more.
Next, the model assumes that all agents can choose from all types of modes. In reality this is not true, as people who do not own a car (or a driving license) can normally not opt for driving a car.
However, as is for all models, this model tries to approach reality by giving an abstraction, and when critically taking assumptions into account, it can still tell us some valuable things.
It allows us to:
Estimate the impact of specific mobility interventions on modal split shifts. Will street closures or shared mobility stations result in more bicycles and less cars, or the other way around?
Estimate the impact of specific mobility interventions on local traffic. If street closures and shared mobility stations are implemented across the city, in which specific areas will traffic increase or decrease?
This dashboard can be used by various groups of interest. But mainly, this dashboard serves policy makers, mobility planners and experts, and city planners in general. It can support them in decision making by exploring possible trends in hypothetical mobility interventions.
The main question this dashboard tries to address is:
Two types of specific mobility interventions are considered:
Transition experiments in city streets
Transition experiments are intentional, temporary changes in street use, regulation and/or form, aimed at exploring systemic change towards a 'post-car' city. They can also be coined as street experiments and are represented as street closures in the MATSim model.
Most of the street closures presented in the scenarios of this dashboard take place in residential streets. This consideration was made to ensure that major traffic axes were not closed, causing undue disruption to traffic throughout the city, as the researchers who developed this dashboard were not familiar with the local context of every city.
Shared mobility alternatives
Shared mobility is the shared use of a vehicle. It provides users with short-term access to a travel mode as they are needed. Additionally, they offer more flexible ways of travelling from conventional public transport. The types of shared mobility services explored in this dashboard are cars and bicycles .
This dashboard analyses traffic on the neighbourhood level (corresponding to the hexagonal shapes in the maps). Two aspects are considered. First, impact on traffic flows is examined: where are the interventions causing change in traffic flows? Second, changes in modal split are considered: how are the interventions altering the composition of traffic flows?
Specific questions that this dashboard addresses (non-exhaustively):
What are the impacts of street closures on traffic on the neighbourhood scale?
What are the impacts of shared mobility options on traffic dynamics when installed in the vicinity of street closures?
What would happen when upscaling some of the scenario's?