Quantified Cities
By 2050, cities will be home to two third of the world population and will concentrate most of the wealth. They will need to change the way they operate and become smarter to benefit from this rapide urbanization and avoid chaos. Qucit — for quantified cities — is a startup founded in 2014 that uses data and artificial intelligence to make cities more efficient, durable and livable.
This summer, I took my first vacation in a long time and visited Myanmar. During my trip, between hikes in serene landscapes and meditation times in Buddhist temples, I spent some time in the booming capital, Yangon.
Yangon is a typical Asian boom town. It is home to 5.1 millions people, double what it was 30 years ago. Getting around using its overcrowded public transport system is a real hassle, and those who can afford it travel in taxis. Power cuts happen daily, crippling economic activity. Waste is not collected properly and a large fraction of the economy is focused on providing food and basic products to the city’s inhabitants.
Still, cities like Yangon are vibrant and exciting. Globally, cities are where most human interactions take place and where economies of scale can be achieved to save resources. Living in cities is a way for humans to collectively organize natural resources supply (water, food, energy), waste collection, health care, transport (roads, public transports, parking spaces), space (parks, streets) and even collective properties (like shared bikes, scooters or cars).
Cities also provide the administrative and legal structure to enable positive interactions to take place. And it works: cities are where most of the world’s cultural and intellectual production occurs (for example as measured by where patents are issued.). Almost 60% of the world’s GDP is produced by the top 600 cities, which account for only 22% percent of the world’s population.
During the next 30 years, the global urban population will grow by 65 million people every year. That is like adding the population of France — every year. More than ever, cities are the engine room of the 21st century.
Considering the huge benefits that cities can bring it’s a very exciting time to be working on cities. But there may be trouble ahead.
Peak Urban?
In densely populated areas like Yangon or New York or São Paulo, one of the most scarce resources is space. When everyone wants to use the same space at the same time, you get congestion. At peak hours, you get traffic jams, subways packed shoulder to shoulder, no parking and air pollution.
The sharing economy is certainly a good way to improve the efficiency of cities by pooling excess capacity. But have you ever tried finding a place to dock your Vélib’ in the Marais or a Citibike in Murray Hill on a Saturday night?
The other great feature of cities is the immense opportunity to interact with others. But above a certain population level, interaction becomes a nuisance, not an asset. More interactions between people leads to a faster spreading of diseases and to more aggressive behaviors. I like moving around cities on a shared bicycle but sharing the road with cars and even other cyclists can be a pretty stressful experience.
Given this unprecedented urban growth, the potential benefits of cities might no longer apply. Consider this: the average speed of a car in Jakarta is 4 kilometers per hour — far slower than a pedestrian. Even in rich cities like London, problems like pollution are worrying.
We may well have reached peak urban. If we have, the consequences will be dire, especially considering the impact on the environment. Depending on how well a city’s logistics are optimized, urban growth can be hugely beneficial or hugely harmful in terms of pollution, CO2 emissions and so on.
Quantified Cities
Thinking about the immense complexity of cities and the challenges ahead led me to leave my job as a physics researcher and build a team of data scientists and engineers called Qucit, which stands for Quantified Cities.
We believe that urban complexity can be understood and mapped. Human behaviors and even human emotions to a certain extent can be precisely modeled and predicted, given enough data and the right tools to understand this incredible resource. This can in turn be used to design more livable cities and city services management that scale efficiently.
Let’s take the example of one of our recent projects with the city of Paris and Cisco. The goal was to redesign public spaces for maximum pedestrian comfort using a completely new data driven approach. To this end, the Place de la Nation, one of the busiest public squares in Paris, was equipped by Cisco with sensors measuring pollution, noise levels, pedestrian and vehicles (bikes, scooters, cars) fluxes or connected device density in more than one hundred locations, 24/7. This is the first time an experiment like this is conducted at such a scale.
We mapped the square very precisely (down to the location of every tree) and we built and trained artificial intelligence algorithms to find patterns in the behavior of people associated with feelings such as comfort, stress, disorientation or security. We surveyed 1300 pedestrians about how they felt depending on a few personal characteristics, their precise location, and other dynamic variables like weather, car traffic or even a demonstration — which is not an unusual event on a Parisian square! Surveyed pedestrians can be considered as “human sensors” for comfort or stress. Their answers were used to calibrate models able to predict people’s comfort levels.
The results are impressive. We were able to rediscover the major factors that impact people’s perceived comfort. In addition, we discovered new ones like the fact that people feel happier closer to older trees. But more interestingly, we were able to precisely quantify the importance of each factor. For example, we found that traffic around the major roundabout alone adds 13% to the stress level of pedestrians that stand on the sidewalk at a 1-meter distance from the road.
During the past 10 years, the amount of data produced by cities has grown by roughly 50% a year, and this trend is expected to continue during the next decade, leading to a 50 times increase by 2025.
Today, Place de la Nation is exceptionally well digitalized compared to other places around the world. However, in a few years this will be the norm and by 2025 we will have a dramatically more precise resolution for our numerical representations of cities. We may have 3D mapping of every public space objects, spatial or aerial images with a resolution of just a few centimeters, sensors measuring hundreds of chemicals (pollutants, odors, etc.) everywhere, and many more “senses” for our quantified cities driven by the rapid spreading of personal sensors.
Modelling a city is complex because behaviors are highly interdependent: for example interaction between neighbors depend on the level of car traffic in the street.
Our artificial intelligence algorithms will continue to improve and the availability of more data means we will be able to train more complex models. These models will be able to quantify precisely the impact of the environment on human behavior and the interactions between people, mediated by the environment.
And it will be an important technological brick on the way towards building a complete dynamical simulation of a city.
Operating System for Cities
The next step will then be to use artificial intelligence to optimize city services.
The first problem we looked at in this respect was the availability of bikes and docks in bike share systems. In every large city like New York, London, Paris or Hangzhou where a bike share system has been installed, a major problem is not finding a bike when you need one or a dock to park your bike at your destination. Most often, trucks are used to rebalance these systems. However, it is usually not done in an efficient way. Strategies are based on average historical data, and on the personal experience of the drivers or their supervisor.
The problem is difficult because static logistics strategies do not work: you cannot use traditional software to rebalance the system assuming bikes don’t move! What you need to do is: 1) predict how people will travel, which stations will experience a high demand and lose bikes, which stations will fill up based on all the contextual data available 2) design a dynamical logistics strategy to optimize the bikes redistribution based on the predicted state of the system. That is what we are doing with our BikePredict Redistribution solution, and it can increase the efficiency of rebalancing by more than 20%.
Rebalancing bikes is only one application of what a quantified city and artificial intelligence can do to optimize city services management. We currently are working other exciting challenges like predicting public transport trips one year ahead to plan future investments, on-street parking search time to route drivers to optimal parking zones and cut pollution or even car accidents to improve emergency response.
We think that two possible roads lie ahead for this urban century: a dystopian world of polluted and clogged megacities or a world where quantified cities function better every day thanks to machine learning, AI and the data revolution. At Qucit, we are working on making sure we take that second path.
Thank you for reading this post! It would be awesome if you’d hit ❤ to help others see it too or share it on Twitter @raphaelcherrier @qucit
At Qucit, we are a team of passionate data geeks that is constantly seeking new challenges to tackle and new ways to build more efficient, livable and sustainable cities using data and artificial intelligence.
And by the way, we are hiring!
Thanks to Mathieu Lefèvre for great conversations & careful re-reading.
