Recent research from a Catalan university shows that the flow of pedestrians in cities can be optimised in much the same way as road traffic.
If, like me, you live in a busy city like London, getting from A to B as quickly as possible can become something of an obsession. Fortunately, there are plenty of tools to help. For drivers or cyclists, traffic-aware navigation is your best bet, and if you rely on public transit, you can use a smartphone-based journey planner. But what about those city-dwellers who prefer to ‘take shanks’ pony’ – or, as they are better known to most, pedestrians?
There may well be a solution for them too, and to learn more about it, we spoke to Dr. Manlio De Domenico from the Universitat Rovira i Virgili (URV). Working with colleagues at URV (Prof. Alex Arenas), MIT (Prof. Marta Gonzalez) and the University of Birmingham (Antonio Lima, PhD candidate), he published a paper in a recent issue of EPJ Data Science that looked specifically at optimising foot traffic in Milan, Italy. “In the paper, we developed a routing system that leverages multiple data sources coming from the city,” said Manlio, “This allows us to provide personalised routing options that balance personal preferences of the individual with common interests of the multitudes.”
To do this, they first had to split Milan into a squared grid with 10,000 cells, each measuring 0.055 km2. But to turn it into a realistic mapping system, they needed data… lots of it. So Manlio and his team entered the 2015 Big Data Challenge from Telecom Italia. It granted access to a huge repository of anonymised Italian telecoms data to more than 1100 researchers in 20 countries. Entrants used the database to do everything from mapping exposure to pollution, to defining a city’s ‘happiness index’. But the URV team looked at the total number and locations of calls and texts generated in Milan, over a period of two months. “This gave us the first layer of data we needed to start building the system”, Manlio told me.
For the next layer, they looked to social media usage – 100,000 geo-located tweets, generated over a 30-day period within the city grid, were used to identify specific events happening in Milan. Layer three came from a series of air quality sensors installed across the city. The final layer was generated from a list of crimes manually sourced from newspaper articles over the previous 12 months. The idea behind choosing these particular datasets (events, pollution and crime) was to identify areas that may be considered ‘undesirable’ when defining a walking route – e.g. an event is likely to lead to congestion, and walking through an area with poor air quality could have an impact on health. As Manlio explained, “Ultimately, we set out to develop a system that coordinates individual’s needs and preferences, and maps them into the wider city network.”
To model the movement of individuals through the city, the team looked to the maths behind the interaction of particles in a fluid. Each trip was treated as an origin-destination-pair, where the origin was a location that ‘repelled’ a particle (in this case, a pedestrian) while the destination was one that ‘attracted’ it. Without tracking individuals, it is not possible to know exactly where a trip starts or ends, so the cellular data provided by Telecom Italia was used as a proxy. The information on events, pollution and criminal activity within the city was then used to define potential routes, so that pedestrians could choose to avoid certain areas, if they so wished.
The results of their simulation were pretty clear-cut. The larger the number of adopters, i.e. those people who followed the suggested personalised route, the shorter the individual journey.
Of course, no matter how good this routing system may prove to be, Manlio admits that it will be virtually impossible to get 100 per cent of people to use it. “As you’d expect, we see the best possible outcome – the closest thing to free-flowing foot traffic – when everyone in the city follows the optimised route. But even with 10% usage, there were some nice effects.” Like any good scientist, Manlio sees a research opportunity here “An individual who chooses any other route is behaving like a random particle, so we’re working on an improved version of the system that tries to bring those ideas in.”
Their particle-led approach offers something else too – a potential way to manage congestion in real-time. “The permeability of the city is a measure of how fast individuals flow through the system – just like particles moving through a porous material,” said Manlio, “By tracking that information, critical congested areas could be identified, and ‘congestion alerts’ sent out to users.”
Scale it up
The URV team hope that their routing system could be used as a basis for a smartphone app that is specifically tailored to a user’s needs, but which also minimises any negative effects for the community. It could even be potentially extended to include other variables, such as street gradient for wheelchair users, or the level of cover in times of bad weather. For each trip, users could offer feedback, to expand the data available to the system, thus improving the route optimisation. “In all of this, it’s vital that the data itself is protected, so that it is completely anonymous – this has become a huge priority for all researchers working in Big Data.” Which is, frankly, a comfort to hear!
So what’s next for Manlio and his colleagues? Nothing less than world domination, it seems J “We would love to extend this model to other locations. In theory, it could already be done for New York and London – two cities that offer lots of open data. Five years ago, it would have been impossible to get access to this level of data, but there’s been a real shift in attitude, and research on human mobility will really benefit from it.”