Utopia says hello. A scenario in which autonomous vehicles dominate the roads and warn each other of dangers, and parking problems, as well as traffic jams, are a thing of the past - for most people this sounds like a distant future. In Berlin, this future is already taking place in several locations today. The "Digital Test Field Berlin" in the Reinickendorf district is one of these locations. Within 25 months, technologies for highly automated and networked driving are to be developed and tested in a typical urban scenario. The main objective of the key "SAFARI" project, which was launched in November 2017 and is funded by the Federal Ministry of Transport and Digital Infrastructure, is to create regulatory framework conditions and technical prerequisites for safe and convenient automated and networked driving (AND). A special focus of the project, which is being carried out in cooperation with renowned partners from administration, ICT research and industry, is on the exchange and updating of digital maps. These are regarded as one of the basic prerequisites for safe, efficient and sustainable mobility. High-precision map material should not only provide information about lanes, traffic infrastructure and traffic light system data; it should also be continuously corrected and supplemented by updates from automated vehicles. Thanks to the up-to-dateness and accuracy of this information, automated vehicles should be able to look further ahead than even the most experienced drivers would be able to. Barriers, for example in the event of accidents, could thus be prevented, and it should be able to optimally approach green phases at traffic lights. According to the theory, all this would lead to greater road safety and better traffic performance - especially if the maps were to update themselves in the end with all traffic-relevant static, partially static and dynamic content.
Visual Article: © Uhura Digital für Berlin Partner für Economics and Technology GmbH
The safety and comfort of automated and networked driving is also being tested elsewhere in the German capital. Since the end of April 2017, a digitally connected test route open to all road users is being developed along the Straße des 17. Juni, which leads east of Ernst-Reuter-Platz to the Brandenburg Gate. In the two-year project "DigiNet-PS", which is also being funded by the Federal Ministry of Transport and Digital Infrastructure, players from research, industry, the ICT industry and transport are working as partners to promote automated and networked driving in Germany.
The approach under the overall project management of the "DAI Labor TU Berlin" is as diverse as the collaborators: "DigiNet-PS goes far beyond the traditional view that intelligence can only be built into vehicles. It takes up the development of decentralization and distributes computing power among vehicles, traffic infrastructure (roadside units, RSU) and digitised road objects as well as the cloud," according to the project description. “In the test field we are not only equipping vehicles“, states Dr. Ilja Radusch, Head of Smart Mobility at the "Fraunhofer Institute FOKUS", which operates the event server in which traffic-relevant events are analysed and distributed in the test field, "but also traffic lights, so as to be able to bring the information of the signal phase - how long is green, how long is red - into the vehicles via communication. We use this information as an additional sensor not only to support the image sensors of the vehicles, but also to make the journey itself more comfortable." Cameras, radar and laser scanners, which enable highly accurate measurement of the distance from static and dynamic objects such as other vehicles, pedestrians or cyclists, ensure that the situation is captured in full. In this way, parking spaces, traffic flow and traffic jams can be identified and transmitted to the vehicles. As if that were not enough, the Fraunhofer Institute FOKUS together with the Daimler Center for Automotive IT Innovations is testing the networking of automated vehicles themselves: they should warn each other of dangers, receive information from the control centre and traffic lights, and plan driving manoeuvres together.
For all this to succeed, comprehensive data alone is not sufficient. Automated vehicles must know how to assess their environment and react to situations. In the artificial intelligence of automated vehicles, algorithms and methods of machine learning are responsible for this: with DigiNet-PS, for example, the enormous amounts of data are collected, analysed and annotated using the FOKUS solution "ITEF" (Integrated Testing and Evaluation Framework) and automated labelling software. They serve deep neural networks of the vehicles as a database to derive successful and unsuccessful strategies for action and to apply this learned knowledge later in traffic situations.
The continuous development of the vehicle software and thus the central learning ability of the algorithms is only one of the goals of this complex project. "We are also making the large volume of aggregated data available for the development of novel solutions," emphasizes Prof. Dr. Dr. h.c. Sahin Albayrak, Executive Director of the DAI Laboratory at the TU Berlin. In parallel projects, a "smart parking assistant" is for example being developed. But DigiNet-PS does not only serve the numerous project partners of the Berlin AI scene as a "test world for city traffic of the future". Companies and scientists from all over the world are invited to put their developments and innovations through their paces here in order to make traffic better, faster, more comfortable and above all safer thanks to intelligent mobility.
© HELLA Aglaia
AI - trendsetter of the economy
While the research work at SAFARI or DigiNet-PS is still in full bloom, some innovations of Intelligent Mobility are already successfully in use. For example, the Berlin subsidiary of the lighting and electronics specialist "HELLA GmbH & Co. KGaA“ has for years been developing hardware and software which enables machines to see intelligently. Based on the premise of "a good view and good visibility in every situation", camera-based driver assistance systems are already in use today which record and evaluate the vehicle environment and thus take automatic interactive measures. “Brighter AI", a spin-off of "HELLA Inkubator", is also pursuing the same goal. The Berlin start-up goes one step further with its system: using innovative deep learning methods, it "calculates" poor visibility and weather conditions from the image, such as rain, fog or poor lighting. For example, a true-to-life daylight version of an unclear night shot is reconstructed to replace it and make the photo or video more clearly recognizable.
The car and ride-sharing capital
The great innovative strength of science and industry in the German capital is also reflected in the data-based free-floating car and ride sharing services. No less than three such systems are in use in Berlin. One of these is the "Free2Move" app, which allows all car sharing companies to be booked with a single click. Regardless of whether a car, motor scooter or bicycle is involved - a map shows which vehicle is available nearby. In addition, the customer can compare prices, locations and features such as the number of seats, drive technology and more.
The special feature of free-floating systems is easily explained: unlike earlier car sharing models, the rental vehicle is not linked to a fixed station, but can be loaned and parked anywhere within a defined urban area via smartphone. The charges are made by the minute; fuel money and parking fees are included. The vehicles which can be used are suggested on the basis of data from the vehicles participating in the transport system. To do this, the vehicle must be located: this works by means of so-called floating car data (FCD) procedures, which turn small cars into mobile sensors and software agents. They provide data not only on the state of the vehicle, but also on the surroundings while the vehicle is standing, for example in traffic jams, in front of traffic lights or in a waiting area. However, the collection of data is only one aspect. In order to find out movement patterns of the vehicles, to predict local demand and thus to better plan the deployment and utilization of the vehicles, the data are analysed using algorithms and systems of machine learning. For Stefan Weigele of the “civity” strategy consultancy, the potential of the innovative systems is undisputed: "Free-floating systems have taken car sharing out of its ecological corner and made it accessible to a broad, pragmatic urban milieu.”
The expert also came to this conclusion as head of the first Big Data study on free-floating car sharing. Free2Move and its competitors have a comparatively low turnover of around 14 million euro today and their local traffic impact is low, the study found. For the year 2020, however, civity forecasts a sales potential of around 1.4 billion euro.
"However, the number of systems would have to increase almost fivefold from 30 today to around 140, and price systems and capacity utilization would have to be further optimised," explains Stefan Weigele, admitting that "the global scalability of the business model is exciting”. The latter is evident with Free2Move. After the successful test in Berlin, "motorized comfort mobility", as free-floating car sharing is called in the study, is now available in more than 33 cities in Europe and the USA.
On-demand public transport
The services of "door2door" are not quite so widespread. Not yet. Since former Harvard and MIT student Maxim Nohroudi founded the company together with Tom Kirschbaum in 2012, the company has pursued a noble vision: to make a car-free city or in other words public transport as comfortable as one's own car. The formula for this is "on-demand public transport", i.e. public transport without fixed timetables, routes or stops. For cities and municipalities to be able to design such a system, they must understand where people are travelling from and where they are going: where do commuters choose their own car instead of public transport due to long waiting times or deficient connections? When will the operation of buses or trains become unprofitable during off-peak hours or in sparsely populated peripheral areas? With the innovative door2door technology, AI algorithms are used to collect exactly this data and analyse the most diverse means of transport in the user's environment - from ride-sharing apps to ride-sharing services and public transport - in order to offer the best route to the user’s destination. In addition, demand-oriented shuttle services can be booked via the platform. door2door has already taken into account the fact that autonomous vehicles will be used for this purpose in the future. For example, the platform enables direct communication between the software and the communication protocols of self-driving vehicles.
Residents and guests of the Portuguese capital Lisbon can see for themselves how AI-based mobility platforms work: together with "Via Verde Serviços", a subsidiary of the transport infrastructure company "Brisa", door2door has developed a route planner app which connects different means of transport and offers dynamic demand-based mobility solutions.
By allygator through Berlin
This is just one of a large number of pioneering projects of the company, which is the only European start-up to be a member of the "Center for the Fourth Industrial Revolution" and to drive forward the transformation of mobility. Since February, door2door has been cooperating with the ADAC in Berlin and is testing a demand-oriented shuttle system in the German capital: together, the partners in the "allygator" project are providing 20 minibuses as shuttle services. As with car sharing, passengers can book buses via smartphone apps, while other passengers who want to travel in a similar direction are picked up on the way. All the passengers are dropped off at their desired destinations. The route is calculated using AI algorithms. The price is dynamic: the more passengers, the cheaper. But it is not only the customers who could benefit from such shuttle systems, which will soon also be autonomous. Transport companies could reduce their operating costs by focusing on actual demand.
According to door2door, the on-demand offer will soon be able to close gaps which classical local public transport cannot serve. It is intended to convince motorists who have hitherto avoided using buses and trains. The long-term goal is to reduce traffic and the environmental impact of exhaust emissions by reducing the number of private cars on the roads. In any case, recent calculations give cause for optimism: according to an OECD study on urban transport in Lisbon, as few as 30 taxi buses with space for eight to sixteen people could replace 970 cars in the Portuguese capital. It would not be the first time that an innovation from Berlin has made a significant contribution to bringing Utopia a little closer.