We have been developing autonomous cars since 2006. FU Berlin is the one and only university in the Berlin region (and one of four in Germany) doing this kind of research. The project started during my sabbatical at Stanford University in 2006. We started developing our own car (named Spirit of Berlin) at Stanford and finished with its development at Rice University, in Houston and San Antonio. Our vehicle reached the semifinals of the Urban Challenge in Nov. 2007. It was a great success considering the short development cycle.
The Stanford researchers we worked with during 2006 moved to Google and theirs became the “Google Car”. We moved back to Berlin after the DARPA competition and started the “Autonomos” project in 2009 with funding from BMBF. Our new vehicle, with the name MadeInGermany, was already driving in 2010 and was certified for city traffic in 2011 by TüV Nord. MadeInGermany is a VW Passat equipped with five video cameras, six LUX laser scanners, one Velodyne Lidar with 64 rays, multiple radars in the front and back, as well as a highly accurate GPS navigation unit. The sensor measurements are handled by several embedded computers, which provide their results to the main computer. The control computer has a GPS annotated map of the city and can order a car to drive from any point in the map to any other. The car keeps its lane and follows the traffic rules: it can detect traffic lights, pedestrians, cars at intersections and react appropriately.
MadeInGermany has been driving autonomously in the streets of Berlin since 2011 and has been featured prominently in the news (CNN, BBC, RBB, ARD, ZDF), in documentaries (CNN, BBC, Discovery Channel, Deutsche Welle, Globo), in the radio, as well as in dozens of newspapers. It has been navigating in the Berlin streets and the city highway ever since, sometimes covering stretches of up to 40 Km (from the new Berlin airport up to Brandenburg Gate). The former Minister of Education and Research, Mrs. Annette Schavan, was chauffeured by the vehicle in 2012, and the Senator for Economics of Berlin, Mrs. Cornelia Yzer, drove in the vehicle in 2013. This is one of the projects at FU Berlin with very high media visibility and public recognition.
The Technology Foundation Berlin (TSB) has declared electric cars and the development of the car industry (for parts and electronics) a Berlin priority. The automotive industry is thus one of the “action fields” of TSB within their Transport and Mobility initiative. Three companies in Berlin are the main producers of driver assistants and components for the car industry: Hella-Aglaia, IAV Automotive Engineering, and Carmeq. We have an ongoing collaboration with all of them.
In October of 2013 MadeInGermany was flown to San Antonio Texas, where it was shown driving autonomously. In the same month, it was flown to Mexico City where it navigated the streets of the city center. The major of Mexico City sat as passenger in the car. MadeInGermany is the first fully autonomous car to have been tested in three different countries, each one representing very different driving conditions.
Our autonomous car project represents a real “Leuchtturm” endeavor within the Berlin scientific landscape. It has been featured prominently in the news, in documentaries, and the BMBF keeps a link from its website to our project website.
Interest of the automotive industry
The DARPA race in 2007 was ahead of its time: interest in autonomous driving really took off only after the race, especially due to the activities of Google (with its Google Car) and the development of new kinds of driver assistance systems which have been entering the market since 2007. There are now have video systems capable of detecting the car’s lane delimiters. A warning can be given if the car leaves the lane. Volkswagen is already offering a “lane keeping” assistant which smoothly tries to center the car in the lane, even if the driver is not paying attention. Volvo and Daimler have introduced emergency stop systems that apply the brakes at low velocities, if an imminent collision is detected. Cars that can park themselves are already in the market. “Stop and go” driver assistants that can assume control of the vehicle during traffic jams are being improved. Every German car company has now a team developing what is called “highly automated driving” systems.
Autonomous driving is a major scientific challenge: reliably detecting all kinds of possible obstacles in a street or highway goes still beyond the state of the art in sensor technology (knowing if a bush is “soft” or “solid” cannot be still be done reliably with laser scanners). The machine intelligence needed for robotic driving is still something where we still do not have all the answers. Detecting the intentions of human drivers or pedestrians goes beyond what computers can do today.
In Germany, only a few universities are working on autonomous cars and have the consequent prototypes (in Munich, in Braunschweig, in Karlsruhe, and in Berlin). However, MadeInGermany is the only vehicle being used on a regular basis, sometimes almost daily, in city traffic. Berlin and Palo Alto are the only two cities in the world where autonomous cars routinely navigate the city. When companies such as BMW or Audi show their autonomous platforms, it is done in a restricted setting (highway or parking lot) and as a punctual experiment.
The scientific field we want to advance is therefore the design and development of the robotic cars of the future, capable of being deployed as taxis in any city. We envision a “vehicle cloud” able to provide service to all inhabitants in a city. A computer cloud will optimize the driving routes, will determine where to pick passengers, will manage the payment for the service, and will transform autonomous cars in part of the public transportation system. The real “green cars” are those that can be shared synchronously, at the same time, so that instead of having 1.3 passengers per car, on average, as in London, we can have three or four passengers in every vehicle.
Three major industrial changes identified by the European Commission for the 21st Century are the development of an intelligent and diversified power grid, the emergence of flexible manufacturing, and intelligent sustainable transportation systems. The project we are proposing here goes to the heart of the third challenge. It has to do with the future of the city and the future of the automotive industry within the next 50 years.
We have collaboration with Prof. Radu Popescu-Zeletin and Dr. Illia Radusch from Fraunhofer Fokus towards developing telematic solutions for communication between autonomous and standard cars in traffic (car-to-car and car-to-infrastructure).
At the international level we have the collaboration with Volvo and with the International Computer Science Institute.
- Frontal laser scanner for detecting vehicles on the road,
- Velodyne 3D-Scanner on top of the vehicle,
- Video cameras on top of the vehicle,
- Ultrasound sensors in the bumpers,
- GPS navigation system with submeter accuracy.
Future components and sensors for our autonomous car will be cheap, modular, and reusable on different kinds of robots.
We are working on several subprojects:
- Computer vision for the road. The computer vision should be able to recognize vehicles and the lane lines. We have a prototype for a car recognizer developed with machine learning techniques. In a cooperation with the company Hella-Aglaia we are now using their computer vision development software.
- New method for tracking objects detected with the laser scanner. At the moment of this writing, we detect objects with the laser scanner in every frame but we do not track them more than a few seconds. It is not trivial to track scanned objects (due to occlusions).
- Combination of the computer vision with the data from the laser scanners, in order to implement a “virtual Lidar sensor” with higher resolution.
- Development of a new kind of 3D-Scanner. We used a Velodyne 3D-Scanner during the Urban Challenge competition. This is a very expensive, yet powerful unit. We are developing our own 3D-Scanner using a special mirror in collaboration.
- We are investigating also high-level control strategies for autonomous vehicles based on reinforcement learning.
In the future we would like to have one single 3D-laser scanner and videocameras as the main sensors. We would also like to be able to drive without an expensive GPS system. The navigation system should be like the ones used today in consumer cars. We are working on such a project in cooperation with Hella-Aglaia.
Our vision is that cars will be autonomous in the future. Car sharing at the level of a whole city can be implemented when the car can pick up passengers, deliver them home or to their offices, and then continue transporting other passengers. An optimal combination with public transportation becomes then possible.
That is why we say that the car of the future will be autonomous, shared, and green, since new energy forms will be introduced in a few years. We are working towards that future.
The autonomous project was selected by the European Commission as one of the showcases for European Research in 2010.