 I will start with a short introduction about the place where I've been working in the last three years, MIT's sensible city lab. MIT's sensible city lab physically is located in the School of Architecture in the Department of Urban Studies and Planning. And this is just a random snapshot of the people working at the lab, or less any time there are between 30 to 40 lab members. The point is that the diversity of backgrounds of these people is really impressive, at least to me. I mean, it's the first time that I am in a lab, or less the number of people in this room, with as diverse best ground as computer science like me, all types of engineering, physics, mathematics, design, architecture, social sciences. And this is really a rich environment. Why is that? Because the focus of the lab are cities, and cities are very complex environments. And to really understand and make an impact on cities, you do need to understand all these dimensions. And that's why we put together Carlo Ratti, the founder of the lab more or less 10 years ago, had this idea of bringing together the design and architecture space with the science dimension, which is something that I think is kind of unique of Sensible City Lab. And just to set a little bit of the stage for what we're talking today, these are numbers. What are these sort of random numbers? Two is the fraction of the Earth that is occupied by cities. It's only 2% of the surface. But these tiny amount of land hosts already today more than half of the world's population. And even more impressing is that there are many predicted trends that says that this fraction is going to increase to up to 3% of the world's population. We'll be living in cities in 20 years from now. And this is mostly due to the contribution of developing countries, of course, where the urbanization trends is much stronger than in the developed countries. 75% is the total amount of energy, speaking of energy, that is currently being used in cities. So if you want to reduce the use of energy, cities is where you have to work. And even more if you talk about emissions, 80 is the fraction. 80% of global emissions are originated in cities. So all these numbers are kind of a good justification of why you are interested in cities. Because working and trying to impact cities is where we can really do a lot of changes. And another focus of the lab is in the so-called big data. This is just a funny definition. It's everything that cannot fit in Excel spreadsheet. More seriously, this sort of random number, it's even difficult to say what it is. I think it's five exabytes of data of gigabytes. It's a huge number. It's more or less corresponding to the entire knowledge of humanity that has been produced since we turned from age to humans till 2003. And all this huge amount of information now is being produced in one and a half day. And of course, the trend is going to be even faster than that. And this means that more and more, the way we will be living in cities will be something like what this image represents. So you have a physical space where we move the city of Dublin, in this case, for instance. But more and more, you will also have a digital layer produced by all the information that all the device that we carry with us or that are embedded in the environment will generate. And we, as people, we will be living in between these two layers, in a sense. And the mission of the lab is trying to understand how the interaction between the physical and the digital layer can be exploited, can be used for the benefits of the citizens and the society in general. And now, before getting into two specific projects I want to mention you today, a little bit of motivation. Of course, I thought that, in general, sensible city lab is concerned with this interaction between digital and physical space. And one of the big focuses on the lab, of which I am the sort of supervising person in the lab, is on mobility. Why mobility? Mobility is a prominent factor. And actually, it's one of the reasons why cities were created, in a sense. So the fact that we meet each other, there is the power of the city. The fact that today, now, we are sitting all in this room to meet is important. But it's the consequence of mobility. So we do not want to take mobility out of the picture, because that would be the easy way of solving of the mobility-related problems. No mobility just connected in the homes to the internet. And that's fine. We never leave our homes except for exceptional cases, and that's done. This is not what we want to do, because we do think that mobility is important. On the other hand, at the current stage, mobility resources are used in a very bad way. We are talking over the lunch about the fact that private vehicles are used for less than 10% of the lifetime. For 90% of the lifetime, they are parked. So they are not only useless, but they also use public or private space that is wasted. Think about the value of public space that is just used for parked vehicles. This is totally useless. Even worse than that, when you have mobility in private vehicles and taxis for instance, just to give you an example, there is a number, which is incredible to see how constant this number is across different parts of the world. The average number of people in a car is 1.3, 1.2, 1.4. This number is constant also in the developing world. And this is another huge problem. So how could we try to change this? Why is that? Because of course, with more people living in cities and with the situation of mobility that is already dramatic in terms, for instance, of emissions, 23% of global emissions are related to mobility, as you were mentioning before. So we need to do something. And that's why, for instance, we start focusing our eyes on one important type of mobility, which is taxi related. Of course, the advantage of taxis is that they have been equipped with GPS since a long time. So there are data sets that we can use to understand how they move. And that's why we started analyzing this huge data set form of 150 million trips performing the city of New York in year 2011. The amazing thing about this picture is that this is not a map. It seems like a map, but it is not the map. If you zoom in, you would see just dots. Each of that dot is the GPS coordinate of a pickup or a drop-off point. When you put together all these dots, you see the map of New York, OK? And this is just to let you understand how pervasive taxi trips are in New York. No big surprise. More than that, we ask ourselves the question, how similar are those trips? How many of those trips could be put together and served by a single taxi instead of two? And that's why, for instance, we put together this website, which is called Hubcap, where you can go. You can select two specific points in the city of New York and see how many trips can be where actually performed between these two specific points during the year. This is interesting. That was a tool that we use also to engage people with the problem of sharing. Yet, it doesn't really tell you whether, if you are now in this point and you want to go to the other point, whether you will be able to find somebody else to share your ride with. So what we did, the idea is that if you have trips like this with similar starting points, similar ending points, and similar times that you cannot see in this drawing here, but of course, the trips should start at similar times. You could possibly put them together. And in principle, you can scale up the idea. Of course, in this case, you might also have to change the design of your taxi, maybe a limo. But in principle, you can scale up this idea. To solve this apparently very difficult problem, we came up with a mathematical model that is just a way of expressing the sharing opportunities. The point is that if you think about it, we have to understand which is the best way of combining 150 million trips. Think about how many combinations. But likely, thanks to this idea of shareability network, we were able to actually solve the problem optimally. And so we came up with curves like this. What do you see in this curve? On this part, you see a quantity that we call delay. What is this delay? The fact is that if you want to share a trip, you have to accept that your taxi might make a little bit of the tour because you have to pick up somebody else and drop off somebody else. So you would arrive at destination a little bit later. This is this parameter delay. And it's in seconds. So if you see 300 here, it's just five minutes. I think most of us can live with five more minutes, except if you are going to take a flight in a hurry or something like that. And as you can see, in New York, with something like around two minutes, you already can share almost 100% of the trips. Even more interesting, what we did, because one might say no big surprise. In New York, 40% of the traffic in Manhattan is made of taxi. So no big surprise. Manhattan is a very compact layout, high density of people. So what we start doing, we started randomly deleting trips, simulating the situation in which there is less demand for taxing. And that's what we got. We got a curve. The important point here is the following. This is the average number of trips in a day in New York, more or less, slightly less than half a million. If you divide this number by two, you still have 100% of the trips that can be shared. If you can divide further by two, still, you can share almost 100% of the trips, meaning that there is a huge potential for sharing trips. And even more interesting is this curve here. So what we did, we asked ourselves the question, OK, this is New York, again, is like the heaven of taxi in a sense for sharing, if you think about sharing. So what happened in different cities? So we get our hands on different data sets from Singapore, from Vienna, from San Francisco. And what we discovered was, to be honest, something we didn't expect. The fact that, basically, we discovered is very similar behavior of sharing. And the sharing opportunities were more or less the same across different cities. And so basically, we came up with this idea of trying to model shareability using a mathematical model that takes very simple parameters, like the average speed of traffic in the city, the area of the city, the delay, the delay that I mentioned before, how many people want to wait, and how many trips that they are generated in the city. And the output of this model is how many of these trips, the fraction of these trips that can be shared. And I don't want to, of course, get into the details. But the point is that the model that we developed really can be used to predict shareability. Why is this important? Because this model allow you to predict shareability also for cities for which you don't have taxi data. Because just to give you an idea, to produce those four plots, this one with the solid lines, we needed to do a lot of data crunching. So a pretty long process. With the model, you can actually predict sharing for any city, virtually any city. And this is, we think, a very important point. And also what is amazing when you see this curve here, this tells us something about the way we live. Why? Because if you think about the layout, the architecture of the cities, the road network of the cities, that's dramatically different. Think about Vienna and Singapore. Very different cities. Yet in terms of sharing, if you have the same density of trips, the amount of trips that you can share is more or less the same. This means that when you think about sharing trips, what is not relevant is the layout of the city. What is relevant probably is the way our lives are organized. So the fact that every morning, more or less at the same time, we're more or less to go to the central business district or in other places where a lot of people work. And in the evening, we made our way back home. On one hand, that's the reason why there is traffic congestion. On the other hand, it's also a huge opportunity for sharing. The other big change that you're going to occur is related to the advent of autonomous vehicles. There are predictions. I don't want to use too much of your time, but there are plenty of predictions says that most of the vehicles on roads will be self-driving in 20 years, more or less. Of course, you can find many of these predictions, and you might have different timelines. But that's the point. Self-driving vehicles are arriving. And another point, if you think about an important piece of infrastructure, which are roads, the average lifetime of roads is in the order of 40 years. This means that whatever we build now is going to be used by self-driving vehicles. So that's why important start asking now the questions about what is the impact of self-driving vehicles on the city traffic. We need to understand it now because whatever decision we take about road network now is going to last for 40 years. So one of the first questions that we ask, of course, I mean, this is a very general question and kind of scary. What is the impact of self-driving vehicles on road traffic? My first answer would be who knows. The second answer is, OK, let's try to start from the simplest point, the simplest block of any road network, which is a single intersection. The intersection is an important part of the road network because it's the place, the physical place, where two flows of vehicles with conflicting trajectories have to negotiate the access to a common piece of real estate if you want, which is the road, OK? And so far, that's the state-of-the-art technology that is used in the busiest intersection in cities, which are traffic lights. I discovered that traffic lights have been invented more than 150 years ago, not even for cars, but for horses in London, in the city of London. And as you can see, that's the traffic light. It was originally a man with just a sign that was turned at time to be shown to different directions. Of course, now they are fancier. They are nicer to watch and see. But the operating principle is exactly the same. You have flow of vehicles coming from different directions, and you give the stop and go to different flows at different times. So of course, now there is going to be this big transition, this big evolution of self-driving vehicles, plus the same evolution in terms of the so-called connected vehicles. So vehicles will be more and more able, and already today they are, able to communicate with each other, to communicate with the roadside infrastructure. So the question is, what will happen with driverless vehicles? We will just optimize the operation of traffic lights, so treating autonomous vehicles as modern horses, or can we do something which is dramatically better, which would imply the death of traffic light? So to answer this question, again, we want to be quantitative. We want to give a definite answer. And in particular, we are not the first that we study autonomous intersection. There were other studies. But they were mostly based just on simulations, without any specific comparison, for instance, in terms of safety. So what we did, we put together a framework in which the same safety conditions are satisfied, both for traffic lights and for our system, which we call slot-based intersection, because it's more or less modeled after the operation in airports, when you have to assign the landing slots to airplanes. So the idea, basically, is that when you enter the road, you send the software entity which manages the intersection a request declaring which is your current position, which is your current speed, and where, which is your trajectory, where you want to go at the intersection. And this piece of software collects all these requests and assign individual personalized slots to cross the intersection to each specific vehicle. That's the key. Instead of dealing with flows, we break down the flows into single vehicles. And each vehicle gets a personalized slot to cross the intersection. And also, it gets the time, of course, because in terms of slot, we are talking about a time slot. So you need to be able to control your speed so that you arrive at the intersection exactly when you are supposed to. And in this way, of course, since you know the trajectory of all vehicles, you can orchestrate all these in order to manage the intersection in a much more efficient way. And now you're going to see a video. Before starting the video, OK, what I can tell you is that the number of vehicles approaching the intersection is exactly the same for traffic light and for slot-based systems. So now you will see a side-by-side comparison. And you will have the impression on one hand to have no traffic, and on the other hand to have a lot of traffic. But all what you see, the difference is just in the way you manage the intersection. The fact that the queues, basically, vanish is due to the fact that the slot-based intersection system is much more efficient than traffic light. Why is that? Reason one is the one that I already mentioned. The fact that you deal with single vehicles and not with clothes. Reason two is that if you think about how traffic lights are operated today, they are operated very badly again. Why is that? Because when there is a lot of traffic, vehicles stop. And it means that when they cross the intersection, they cross the intersection with a very slow speed because they start from speed zero. So basically, in terms of making the system more efficient, you would like that the intersection area should be used by vehicles for the shorter time possible because after that, you can give that area to other vehicles. And now, with current managing of traffic lights, this is not the case because basically, the average speed at which vehicles cross the intersection is very low, probably five to 10 kilometers per hour. With this system, you can slow down vehicles before arriving at the intersection and let them cross the intersection at the perfectly safe speed, like 30 kilometers per hour, yet much faster and so that you can serve more vehicles. And now, just to finish my talk, challenge is related to transition to self-driving. I mean, these are kind of the collection of the typical questions when we make our presentation. One is what happens to pedestrian bikes. Of course, one of the reasons of that is because in our video on purpose, we just show vehicles. But what I can tell you is that the system is very flexible and allow you, for instance, to reserve for pedestrian, to be integrated very easily. You can think about having an app. You can just press a button to request a personalized slot to cross the intersection. And you would have on your smartphone a countdown to let you reach the intersection at the right point, the same for bikes, just to give you an example. Of course, there is, again, a matter of opportunity. Do you want to do that or not? The reason is, since pedestrian and bikes are much slower than vehicles, you would slow down the operation of the system. This is perfectly fine. This is just an engineering. If you want a more political decision, we as computer scientists and engineers just give the tool to manage the intersection, which is the best way, in a sense, is a choice of the city authorities or regulators, traffic authorities, and so on and so forth. Another big is user acceptance, something that you talk about during the lunch. In this image, you see people that basically, as you can see, that's the direction of travel. And they're just basically sitting backward watching TV. I don't know how quickly this scenario is going to turn into reality. But of course, there is a lot of work to be done. I think, as we were discussing over the lunch, that is going to be an incremental process. So we will get more and more use to partially self-driving features like, I don't know, a sort of autopilot in the highway, something like that, or this feature like automatic parking, in which you just get off board and you press a button and the vehicles go and look for the parking spots. When you get used to that, you will have more and more confidence in your vehicle. And maybe you can eventually get into one of these conditions. Of course, other big challenge is the ownership model. Would it make sense at all to own a vehicle when it's self-driving? It can come pick you up any time when you want it. Probably a different vehicle every day, because sometimes I do this exercise with myself asking, which is my ideal vehicle? Because if you think about it, and other problems that currently we tend to buy our vehicle based on the highest, largest possible use. So if occasionally you need for a long trip a vehicle with a lot of trunk, then you buy a vehicle with a lot of trunk. And 90% of the time you use it only you. So again, it's kind of a waste if you think about it. So ideally, if you can say the evening before for the following morning, tomorrow I would need at 8 o'clock a three-seat vehicle, fine. Why should I own a specific vehicle? But again, this is also a societal change because there is a lot of more than just mobility associated with car ownership today. And finally, the other big issue is liability. Who is going to be liable for the hopefully less likely accidents? So my guess is that my belief is that there are going to be much less accidents, but there will be accidents. And those accidents would be much heavier than the infinite number of accidents that occur today with human drivers. It's more or less the situation with airplanes, right? The airplane is by far one of the safest transportation modes, yet when there is a single air crash, it's a big news. Millions of people dying on the road, nobody cares. So that's another big challenge. So to summarize in a single slide what I think, what we think would be a possible future of mobility, we think is going to be something much more flexible. So the current situation is that the mobility landscape is very well divided into well-known categories like the private vehicle, the taxi, the bus, the train, right? But in the future, there will be a continuous probably of transportation modes thanks to this conversion between the so-called sharing economy, so the fact that people want to share more and more resources, including mobility resources, and also autonomous vehicles can play a big role, right? Also in sharing vehicles, because one of the problems, for instance, of sharing a car is that you need to rebalance at some point, because it's something that, for instance, for bike sharing companies, it's a cost that at some point there should be a van that takes bikes from one station and bring the bikes to another station, OK? If you would have self-driving, probably you could solve that problem. And in the meanwhile, the vehicle could do something useful in between.