If someone asks what fog computing is during a conversation, it is likely that the rest of the group will be left in silence with questioning faces. On the other hand, if the suggestion is to imagine what cities will be like in one hundred years, there is no doubt that self-driving cars will be the first elements to bring their ideas to the table in the debate.
We chatted about the “Internet of things” with the fellow Damián Roca who has recently been awarded one of the Computing Awards of the BBVA Foundation and the Spanish Computer Science Society for his research into self-driving cars.
Q. What is “fog computing” and what do self-driving cars need it for?
A. The future of self-driving cars will be about managing to develop artificial intelligence systems so that they can travel error-free according to the environment and based on experience and learning, in the same way that people learn to drive.
The technology that I have developed is a stepping stone between the current situation with these types of vehicles and the end goal. Right now, we have self-driving cars such as Tesla vehicles which have individual sensors that collect, process and send data to servers which can be anywhere in the “cloud”.
With fog computing what we do is send and process the data to nearby places where it is generated via the creation of a network of nodes (devices capable of communicating with other devices, storing data and processing it), thus making it possible to reduce the response time for the processing. For example, we have a car travelling along a street where there is a red traffic light, if we apply this technology we can place a device on the traffic light that directly sends the signal to the car and tells it that it must stop. It is a way of directly connecting the vehicle to its real environment and reducing the waiting time for data processing.
And, if we want to take it a step further, we could end up with self-driving cars not even needing that traffic light device to know what they need to do by applying rules similar to those that birds follow when they fly in groups.
Q. “Fog computing” could really change people´s lives... Is it the first step towards radically changing the configuration of cities?
A. Yes, it is an important step towards obtaining change in our cities and moving towards smart cities as fog computing would help to optimise infrastructure. That is to say, the nodes for processing, storage and communication installed in a city, which would be used for the self-driving cars, could also be shared for other applications, for example, for public transport data.
While we are still talking about self-driving cars, my research used fog computing technology as a starting point to attain another result based on the observation of how flocks of birds behave, which would be the next step: Hierarchical Emergent Behavior.
If we watch a group of birds flying together we observe that they always follow the same dynamic: they never hit into one another, they all move close together and adjust their speed and direction between themselves. This is a biological concept that, along with the concept of hierarchical organisation provided by engineering, offers us the key to organising a group of hundreds or even thousands of self-driving cars travelling around a city.
It would involve applying these same rules at a local level in order to organise the traffic in a city, complemented by a network of nodes close to this large fleet of vehicles. So, continuing with the example of the flock of hundreds of birds: if you are one of them, you would have to calculate data for distance, speed, etc., in relation to everyone and the amount of data would be enormous. On the other hand, think about how that volume would change if you simply needed to modify your flight pattern by focusing on what the six or seven neighbours flying at your side are doing, which is what birds do in flock.
This is what we achieve by applying the rules at a local level, we would not care about the total number of cars there are in a city and we would simply need to focus on the actions of those that are nearby in order to modify our behaviour.
Q. In many cities in the USA it is now common to see electric cars on the road, when do you think that will happen in this part of the world?
A. I think that what we need now is to be able to charge electric vehicles and to do so, there needs to be an energy distribution network. For example, in the Silicon Valley area, there is a very extensive network of chargers offering the same characteristics as petrol stations. Here, in this part of the world, it is a lot less developed but I think that when this is resolved, electric cars will arrive with a lot of momentum.
In terms of a time scale it is hard to calculate because we have a lot of regulation and legislation that affects energy and transport, but I think that within ten years we will begin to see a higher presence of these vehicles. Getting that moment to arrive will also be linked to different commercial alternatives, that is to say, not just having a model but getting all the brands to begin to sell them and getting people used to electric cars with the changes they entail.
Q. It seems that we are still a long time away from living in smart cities, how much time do you think we need?
A. Electric cars are a first step but, in order to implement the networks of sensors and actuators (devices that can activate mechanical elements) that make it possible, for example, to automate transport, for everything to function with clean energy, for citizens to have lots more services or be able to link up more cities within a city, that is going to take quite a lot more time because the companies involved need to reach an agreement.
If we look at the different systems that we might have at home, we see that now Google sensors or tools only communicate with Google and its associates, and the same goes for the rest; that is to say, there are no general standards that can be shared.
These companies have their businesses in different fields and standards and homogenisation are needed in certain areas, for example in data exchange and defining how new applications are translated for citizens.
On the other hand, there are also other sensitive issues that will need solutions, for example, data processing and protection as there will be people who do not want to put their routines and personal information into the hands of companies with dubious objectives.
Q. At present, “big data” is the epicentre of any business, will it be viable to change to the scale that your technology proposes?
A. Big data entails a change in the sense that many years ago the problem was accessing information and at present the problem is that we have access to too much information and the difficulty is distinguishing between the valid part and the false part and “noise”. There are currently few companies that can afford to process big data but it is true that this information is like the “new oil”, ultimately information is power.
That is to say, twenty years ago if you did a university degree and needed a piece of data you went to a book to search for it and now you enter Google and it instantly shows you thousands of results. The problem is that at a human level, although you have a great advantage in having access to that information, it is very difficult to process and it takes a long time to extract what you really need. The advantage of big data is indeed doing that work in a short span of time and more precisely.
Going back to the example of the self-driving car, right now a single vehicle captures the video of the route it is following, the data it detects and generates from the sensors inside the car, that of the people travelling in the vehicle, interaction, etc. A very large amount of information is generated and now it is sent to servers that are very far. The advantage of our technology is that by sending and processing this data closer, we take less time to process it and also, the result is closer to reality. If we also apply the “three rules of birds” then what we do is get the vehicle to make its own decisions based on what it observes.
If we think about a car that is guiding you using Google Maps, currently the route is calculated by a server that is not very close and nor can you see the traffic around you. With fog computing and the three rules, you don´t need to wait for the server to detect a traffic jam and then send the signal that you need to reduce your speed, instead, the vehicle processes the data in real time more precisely; all of this encourages big data because we are simplifying and sending “cleaner” data that is fully processed.
In any case, what fog computing technology changes is the infrastructure that enables the traffic, processing and storage of data on devices, therefore the end users are not really going to note a lot of difference beyond a good user experience. It is like when you make a phone call, you don´t know how many antennas your call is going through, but ultimately you will know if there have been interferences, if the connection has been lost or not and other details like that.
Q. In addition to the automotive sector, what other fields can “fog computing” be applied to? Are any of them particularly revolutionary?
A. Fog computing can help a lot of other fields in addition to the automotive sector. For example, the one with the most impact right now is smart cities and everything that “fits” into them such as smart homes, and transportation systems. Within this environment, fog becomes the platform that all of the different systems of a city are implemented on and the medium where data can be exchanged.
To give another example, it could also be used to control the work of robots in a car factory, or in the case of the dynamic of a city, to receive and process information about the position of people in order to find out where there is most demand for public transport in real time.
Q. Have you finished your research into “fog computing”? What is the next step? Could you explain what new projects you are embarking on?
A. My personal research is at a bit of a standstill as I am focussed on a new venture, but a very large community has been created around the areas of development of fog computing where there are lots of people contributing, lots of people testing and validating in different areas. I think that the next step in this research is to achieve a scaled-up deployment, that is to say, to get an entire city to install the networks of sensors.
And I have set up a company in Spain based around my new projects. Its name is Tocat Labs and I work with Consilient Labs Inc. which is in the United States, as our idea is to have a presence in both countries.
We are still at the early stages of funding and development and we can´t explain too much but we are working on a new method of machine learning based on not getting the algorithm to memorise a pattern, but rather getting it to understand, reason and reach conclusions.
Photography by Damián Roca: Manu Mielniezuk