AXS: An Agent-Based Simulation Model for Urban Flows

PhD Research

Although ephemeral, urban flows can have a lasting effect on city organisation, affecting decisions taken by people regarding where to live, set up business and carry out activities. Thus, flows can be identified as an important factor shaping urban form while being shaped themselves by a series of urban and environmental factors. These include: land uses, which have an effect on trip generation by setting origins and destinations; infrastructure and terrain characteristics, which can make it easier or harder to use different routes; cultural aspects and collective memory, which can make people prefer one route over another, and so on.

My PhD research proposes an agent-based simulation model, entitled AXS, which aims to replicate urban patterns, such as the ones generated by accessibility and centrality measures, from the bottom-up by simulating the movements of individuals in cities. Thus, this research aims to simulate flows from the bottom-up, but to analyse the results on a larger scale and aggregated form. The proposed model is capable of generating artificial flows constrained only by urban morphological properties, but it can also be extended to simulate more complex movement patterns by differentiating agents’ behaviours and preferences, by generating trips based on real land use distribution, by including different transport modes, and so on. The flows patterns produced by the AXS model are compared to patterns of accessibility and centrality measures, as well as commercial and residential land uses distributions and traffic flows in real cities. The model allows for the disaggregation of the study of flows using different types of groups (such as socio-economic or ethnic groups) as well as to study the relationship between urban morphology and flows of movement.


The first experiment developed with the AXS model was carried out for the study area of the Greater London Authority (GLA) area. This experiment simulates urban flows with randomly generated origins and destinations inside the urbanised area of the GLA.

The second experiment developed for the GLA study area was carried out with real flow data from the UK 2010 Census. The urban area was divided in zones (MSOAs, from the census) and the information on the number of people living and working in each area was input to the model. A sample of those flows was simulated, and the result can be seen in the video below:

In the AXS model, flows can be disaggregated by population groups defined by characteristics such as ethnicity, income, occupation, etc. This can give us insights into the collective ‘activity space’ of each group, highlighting the more diverse and the more segregated areas of the city in terms of flows. The video below shows the results for the GLA area, disaggregated by ethnic group:


An application of the AXS model to the study of segregation was presented at the ABEP-UK Conference 2018, which took place at London on the 10th of March 2018.

This research was presented at the American Association of Geographers (AAG) Annual Meeting, which took place in Boston, USA from the 5th to the 9th of May 2017.
abstract | presentation