Well behaved cities – what all cities have in common

By: Dyrol Lumbard, Mathematical Institute, University of Oxford

picture of ancient city

How are people, infrastructure and economic activity organised and interrelated? It is an intractable problem with ever-changing infinite factors of history, geography, economy and culture. But a paper by Oxford Mathematician Hyejin Youn and colleagues suggests “a mathematical function common to all cities.”

Think of the city as an ecosystem, types of businesses as species interacting in that system. Ecosystems in the natural world often share common patterns in distributions of species. That got the researchers thinking. Maybe the same consistency arises in the city too. Only instead of the food web, it’s people and money and businesses that require one another. We usually think of cities as unique. London is very different from Moscow. But, it turns out, what governs the distribution of their resources stays the same across the board.

The team analysed more than 32 million establishments in U.S. metro regions. An establishment, the unit of analysis of their study, indicates “a single physical location where business is conducted”. When the team measured relative sizes of business types (e.g. agriculture, finance, and manufacturing) in each and every city, and compared these distributions among cities, the universal law is found: despite widely different mixes of types of businesses and across different-sized cities, the shape of these distributions was completely universal. Cities have their own underlying dynamics. It doesn’t matter where they are, how old they are and who is in charge.

This underlying pattern allowed researchers to build a stochastic model. As cities grow, the total number of establishments is linearly proportional to its population size (more people, more businesses). When an establishment is created it differentiates from any existing types with a probability which determines how diversified a city is given its size. This probability turns out to be inversely proportional to city size: the more businesses, the harder it is to differentiate them from existing businesses. This process, with further research, displays an open-ended, never-ending, albeit slowing, diversification of businesses in a statistically predictable way, constituting a human eco-system.

For a fuller explanation of the work also see articles in Forbes and Next Cities.


This article was first published on the website of the Mathematical Institute, University of Oxford, on 22 February 2016.

Are big-city transportation systems too complex for human minds?

By: Dyrol Lumbard, Mathematical Institute, University of Oxford

mapping of transport routes

Many of us know the feeling of standing in front of a subway map in a strange city, baffled by the multi-coloured web staring back at us and seemingly unable to plot a route from point A to point B.

Now, a team of physicists and mathematicians has attempted to quantify this confusion and find out whether there is a point at which navigating a route through a complex urban transport system exceeds our cognitive limits.

After analysing the world’s 15 largest metropolitan transport networks, the researchers estimated that the information limit for planning a trip is around 8 bits. (A ‘bit’ is binary digit – the most basic unit of information.)

Additionally, similar to the ‘Dunbar number’, which estimates a limit to the size of an individual’s friendship circle, this cognitive limit for transportation suggests that maps should not consist of more than 250 connection points to be easily readable.

Using journeys with exactly two connections as their basis (that is, visiting four stations in total), the researchers found that navigating transport networks in major cities – including London – can come perilously close to exceeding humans’ cognitive powers.

And when further interchanges or other modes of transport – such as buses or trams – are added to the mix, the complexity of networks can rise well above the 8-bit threshold. The researchers demonstrated this using the multimodal transportation networks from New York City, Tokyo, and Paris.

Mason Porter, Professor of Nonlinear and Complex Systems in the Mathematical Institute at the University of Oxford, said: ‘Human cognitive capacity is limited, and cities and their transportation networks have grown to the point where they have reached a level of complexity that is beyond human processing capability to navigate around them. In particular, the search for a simplest path becomes inefficient when multiple modes of transport are involved and when a transportation system has too many interconnections.’



Professor Porter added: ‘There are so many distractions on these transport maps that it becomes like a game of Where’s Waldo? [Where’s Wally?]

‘Put simply, the maps we currently have need to be rethought and redesigned in many cases. Journey-planner apps of course help, but the maps themselves need to be redesigned.

‘We hope that our paper will encourage more experimental investigations on cognitive limits in navigation in cities.’

The research – a collaboration between the University of Oxford, Institut de Physique Théorique at CEA-Saclay, and Centre d’Analyse et de Mathématique Sociales at EHESS Paris – is published in the journal Science Advances.


This article was first published on the website of the Mathematical Institute, University of Oxford, on 19 February 2016.

Creating a Smart Parking system in Oxford using Nominet’s IoT Tools

By: Adam Leach, Nominet

As mentioned in a previous blog post, we have been working on a new application – smart parking – to test our IoT Tools. Using Nwave’s car parking sensors (which use the Weightless-N wireless protocol that Nominet contributed to), and our IoT Tools we have been able to quickly build a system that allows us to monitor car park usage at Nominet’s Oxford HQ in real-time. In this blog post we explain how we used the tools to rapidly build this new application.

Although the IoT tools were initially built for the Flood Network, they were designed from the outset to be flexible enough to work with any IoT application, so the smart parking project was a great opportunity to test that theory. On the surface, flood monitoring and car parking don’t appear to have much in common. However taking a step back there are consistent challenges; both applications require time and location-based data to be collected from sensors, processed, analysed and then presented visually – exactly the tasks that the IoT Tools were built for.

car park

As with the Flood Network, one of the most challenging parts of building the smart car parking application was physically installing the sensors. However, once these were in place the rest of the process was straightforward. Configuring the car park sensor devices (and associated data streams) was a very similar process to configuring Flood Network sensors. The challenge here was to make the configuration process as quick as possible for the deployment of more than 100 devices in one go. To assist with this task we designed the Management UI to accept plug-ins, with one of these being recipes. We created a recipe that can import a CSV file with a list of device parameters (e.g. ID, name, location) and a template device. A template device is a device that has been previously defined and configured in the IoT Registry. Using this recipe it was a matter of setting up one device and replicating the same configuration for all the other devices using the parameters from the CSV file. The whole process took minutes.

Once the car park sensors were configured in our system we could immediately use the data monitoring, visualisation and analytics available in the management UI. For example, we used the time series data visualisation tool to help with verification of the magnetic calibration process used by Nwave’s sensors.

data monitor

For the map visualisation, we already had a generic framework for displaying sensor timeseries on a map, and styling their associated geo-objects. Expecting that each project would have specific needs, we made it easy to plug in custom data visualisations and interactivity. For the Flood Network these timeseries are river levels, the geo-objects are the rivers themselves, and the data visualisation is the sparkline graphs.

car park diagram

Car parking is a very different problem from river flooding, but the core issue is still time-domain measurements of a geo-object; in this instance, the geo-objects are parallelograms representing the parking spaces, and the timeseries are their occupancy over time. Our framework was already a very good fit. All that needed to be done was the custom data visualisation; the ‘ribbon’ plot for a single space over time was easy to implement in D3 (an industry-standard library), and so was the sparkline representing total occupancy. Add in the click/touch/hover interactivity, customise the look’n’feel, and we’re done – real time car parking occupancy viewable on a map.

So what’s next? We are continuing to look at new applications to further expand the functionality the IoT Tools beyond purely sensor measurements, and are currently working with partners in the education and conservation space (more on this soon). If you are working on an IoT application and think that the IoT Tools could be of use to you too, then please get in touch at adam.leach@nominet.uk.

– Adam

This post first appeared in the Nominet R&D blog on 1st Feb 2016.

Smart cities and collaborative mapping tools

By: Jonathan Bright, Oxford Internet Institute

The UrbanData2Decide project has partly been about getting to know local government administrators and understanding more about the types of data related challenges they face when policymaking. One common refrain is that there simply isn’t enough information, and what there is goes out of date quite quickly. Hence decisions have to be made on common sense, instinct or ideology. One of the beliefs guiding the setup of the UD2D project was that social media data might be able to ameliorate this situation.

 heat map of city region

In particular, one project we’ve been working on concerns the use of point of interest data coming out of OpenStreetMap as a way of providing a picture of the urban environment (for example locations of venues selling alcohol). Collaborative mapping tools such as OSM can offer a very rich picture of local life, as they are created by people who live in the local area.

A first task in the use of this sort of data concerns its validation – how accurate is it? Hence what we’ve been trying to do, looking at OSM data, is determine how complete it is, and how this completeness varies by area. You can see such a validation applied to Cambridge above (darker zones are more “complete” mappings). Clearly there is important variation in how complete the data is, and this variation isn’t randomly distributed (seems to be more complete in the city centre). In later work, we’ll be looking more systematically at what explains this type of variation.

NB This work was produced in conjunction with Stefano de Sabbata while he was a fellow at the OII.  The article was first published in February 2016 in the Smart Cities Research Blog of the Oxford Internet Institute.