How to get Level 5 autonomous driving on the road by 2019
Tele-operated driving has long been known as a research area. Now, the development of driverless cars could speed up – thanks to interaction with the related infrastructure.
Be warned: Should you find yourself at the Mercedes-Benz Museum in Stuttgart, you might meet an unmanned vehicle coming the other way in the car park. Early next year, Mercedes and Bosch plan to launch a pilot project with the intelligent parking guidance system. The year after that, the launch of the system could be one of the first Level 5 applications worldwide. So how will it work? Using an app, test clients can reserve a vehicle that will start driving autonomously at a requested time. Returning the vehicle is just as easy: users can leave the car in a drop-off zone and complete the process via the app. While the passengers stroll around the museum, the car slowly cruises around the car park looking for a place to stop.
So, could driverless cars be going into mass production sooner than even the experts predicted? “It is still true that we will not have robot taxis ready for mass production before the beginning of the next decade,” says Gerhard Steiger, Chairman of the Chassis Systems Control division at Bosch. Until then, test fleets of autonomous cars ought to gather experience in all corners of the world. Daimler’s Smart presented suitable prototypes at the IAA in Frankfurt in September. For private customers, fully autonomous vehicles will not really be affordable at first. For robot taxis, on the other hand, a one-off investment in sensors and supercomputers will play a much smaller role. Quite simply because taxis constantly earn money.
Fewer costs, quicker implementation?
An inexpensive alternative are systems based on the infrastructures that outsource the vast majority of required sensors and computing power. “When it comes to automated valet parking, the car itself changes very little: in addition to electronic steering and braking, only one communication module is required. This module receives commands from the car park’s central computer and transmits them to the engine’s electronics,” explains Steiger. “These systems are known technical solutions which are already on the market. They mean fewer extra costs for the car and can be widely implemented, much faster.”
“Fast” is of course a relative term. The automated car park idea first came about in 2011. In a research project with the Fraunhofer Institute for Open Communication Systems in Berlin (Berliner Fraunhofer Institut für Offene Kommunikationssysteme), Daimler developed a navigation system for car parks known as “eValet”. Using Car2x communication, a central computer directed the cars through the “maze”. Admittedly, with a human driver behind the wheel. But even back then, with the help of cameras in the roof, the system could recognize the exact position of the vehicle and direct the driver to the nearest available space by showing instructions on the central display. “I’m amazed that it took 6 years to develop a product from our project,” says Dr. Ilja Radusch, Director of the Fraunhofer Institute, “but the devil is clearly in the detail.”
A control center for autonomous vehicles
Just like fully autonomous vehicles, intelligent infrastructures also have to solve a number of problems: Is the white object in the car’s path a plastic bag or could it be a child’s diaper? Currently no one dares to answer the question without the use of LIDAR scanners in the car park. Only in the next phase are significantly cheaper cameras set to take over the monitoring of the environment. So can such an approach be applied to traffic flow? “Fully autonomous cars have to manage without a driver in all conditions, such as wind and rain. Perhaps what is required is a control center that can be activated in critical situations,” says IIja Radusch.
Tele-operated driving certainly requires a number of infrastructural measures. On the A9 motorway in Germany, Vodaphone, Bosch and Huawei are currently testing a specially designed high-speed mobile network. Many small radio cells allow real-time communication between autonomous vehicles. Within a distance of 320m, information about speed, position and lane changes can be constantly exchanged. The sensor measurements can also be sent to the Cloud to warn others of potential hazards such as ice or aquaplaning. Yet with a conventional mobile network, this is far from easy. Particularly as control centers now need real-time access to all sensor data from an autonomous car. For example, to remotely give a robotic taxi permission to keep driving.
Project “Inframix”: simulate and test traffic flow with driverless cars
The Fraunhofer Institute for Open Communication Systems recently presented a further research project called “Inframix”. It will investigate the effects of autonomous vehicles on traffic flow and road infrastructure. The project is being supported by the European Union until 2020 with 4.9 million euros. Project partners BMW, Siemens, TomTom and Austrian public road infrastructure cooperation ASFiNAG are matching this amount, thus doubling the total project budget. “Almost 150 experts will work on the project. They plan to test traffic flow with autonomous vehicles – first in computer simulations and then with traffic guidance systems in real life,” explains Ilja Radusch.
Simulations can test an empty robot taxi for example – and their impact on traffic. Ilja Radusch also expects such demo projects to be of great importance to authorities and (municipal) politicians: “Autonomous driving is still quite a new subject for most mayors. We need to be able to present different scenarios with a simulation tool.”
Ultimately, the extent to which tele-operated driving is necessary and possible needs to be defined. Nissan has already announced that they want to connect their autonomous vehicles to a control system. However, a manufacturer-exclusive system is unlikely to have a future. This is why the Inframix project is designed to develop a “hybrid” road infrastructure. Both conventional and autonomous vehicles should be guided at the same time in various traffic scenarios, such as dynamic lane allocation for highly automated vehicles or road narrowing due to construction sites. Real-life trials in Austria, Spain and Germany will conclude the project. The results could be significant for robotic vehicles around the world.