At the Electronics Research Lab in Silicon Valley, experts from a variety of disciplines are engineering the future of mobility.
Steffan Heuer (copy) & Benne Ochs (photos)
Steffan Heuer (copy) & Benne Ochs (photos)
The Audi concept car moves unerringly across the car park in front of an office building in Silicon Valley and keeps precisely to the center of the lane despite the lack of markings. Its three cameras detect intersecting traffic lanes to the left and right as well as parking bays, enabling the vehicle to cross junctions safely and turn off effortlessly by itself. What is really striking is not the fact that the driver has taken his hands off the steering wheel and allowed the vehicle to drive itself.
No, this demonstration of piloted driving is remarkable because this Audi has learned within a matter of hours to master a task hitherto unknown to it. Based solely on images of its surroundings that are captured by conventional cameras and fed into what is called a deep neural network, the algorithm employed here is capable of making decisions about the planned route, and learning from each successive journey. “This is an amazing capability: a car that can learn how to drive — just as a person learns something new and improves steadily over time,” says Ewald Gössmann, executive director of the Electronics Research Lab (ERL) at Volkswagen Group of America, south of San Francisco. “This approach does away with stored navigation maps, position finding, and pre-programmed software.”
Experts describe this as “deep learning,” a system of adaptable algorithms that becomes increasingly capable and intelligent over time, just like the interconnected cells in the nervous system of a living organism. “The technology involved is compact, fast and affordable enough to use neural networks in dealing with the many and complex aspects associated with automated driving,” adds deep learning expert Lutz Junge, whose team brought the intelligent system to road-trail readiness in a matter of months. The ERL is part of an international, inter-company network, and comparable think tanks organized along similar lines are up and running at the Audi headquarters in Ingolstadt. “For us, the key thing is networking these embryonic ideas and using the data and concepts collectively. The pieces of the puzzle from which future interconnected functions will be created and the new development tools are so diverse that sharing the ideas and insights gained, not to mention the cooperation across locations, will become key success factors,” explains Klaus Verweyen, head of pre-development automated driving functions. For example, expertise from the ERL and from Audi is incorporated into a functional architecture in the Audi development vehicles already taking part in piloted driving trials on freeways or racetracks today.
At the ERL, groundbreaking achievements are almost an everyday occurrence. Since it was set up in 1998, the laboratory has become an integral part of Silicon Valley and currently employs almost 200 experts across a wide range of disciplines. They work to create elegant technical solutions to meet the needs and wishes of today’s customers for the seamless, convenient and safe mobility of tomorrow. Virtually all of the meeting rooms, named after Californian landmarks, are equipped with large glass walls on which inspirations can be quickly sketched. The space between workstations is strewn with everyday consumer gadgets offering an open invitation to experiment: fitness trackers, networked loudspeakers, wifi weighing scales, interspersed with various prototypes created using 3D printers on the ground floor — all of which point to the incredible pace of work at this laboratory. “For us, the customer is at the front and center of our thought processes,” explains Gössmann, who has headed up the ERL since 2014. “Only when we know what people really want does the technology come into play. This approach lets us create an experience that enriches digital mobile life for the Audi brand.”
In this regard, Gössmann and Chuhee Lee, director of technology and strategy at the laboratory, are concerned with much more than just vehicles and the hardware and software they contain. “Mobility is a basic human need with numerous facets. Every journey begins long before you sit behind the steering wheel, and ends long after you exit the vehicle,” explains Lee, who has spent the last 15 years in Silicon Valley scouting the latest trends. For him and for Gössmann, mobility includes things like wearables and elements of smart home technology up to and including networked light bulbs, which, of course, use LED technology. “The car is merely a part of this eco system,” adds Lee. “Our vision is to create an end-to-end experience.” Put simply, the goal is to build systems that know people and adapt to their needs.
To achieve this aim, the ERL has put down deep roots in the Valley’s creative process, cultivating research partnerships with leading universities such as Stanford and Berkeley as well as maintaining contacts and memberships in start-up networks and expert panels like the Singularity University. It also works with leading technology partners — for instance, Google and Apple — and with well known venture capital firms. Companies that set their sights on pinpointing ideas and tapping their potential ahead of the posse must also think and act internally more like an agile start-up. This is why the ERL brings together engineers, programmers, psychologists and designers who in many cases started careers with companies not linked to the
Input is always in demand. Since the start of 2016, the ERL has operated an in-house incubator in which small teams of employees can develop and nurture ideas. The ideas that gain traction in turn gain access to budgets typical of a Silicon Valley start-up. In the first four months of its existence, over 20 innovative projects were initiated in this way. “These projects are continuously audited, adjusted and, in some cases, abandoned,” reports Lee, “but even that is an important experience for us because it establishes a dynamic feedback culture.” In this way, new platforms that can be expanded with more functions are developed. One example is the self-learning “Personal Route Assistant,” which is also used in the new Audi Q7 and Audi A4. Once activated by the customer the system learns the driver’s regular routes and destinations. It then links this information to the vehicle’s current location and time of day. So the system learns from the driver’s behavior, using what it has learned to suggest optimized routes on the next journey — even if route guidance is inactive.
The navigation system considers the three most likely destinations in its calculations, which also take account of the arrival time and current traffic conditions. For instance, it can suggest that the driver activate the system to find potential alternative routes. The customer can activate or deactivate the service at will or delete destinations stored by the service. Deleted destinations as well as routes driven by the customer with the service deactivated are neither stored by the system nor taken into consideration. “The vehicle adapts to the driver, not the other way around,” says Gerardo Rossano, who, together with Sagnik Dhar, developed the system from an academic experiment through various test stages all the way to market maturity. “The ideal scenario is where my smart home already lets me know in the morning whether anything new has come up and suggests a time of departure,” says Rossano.
Gerardo Rossano (right) and Sagnik Dhar (left) are developing a smart navigation system that adapts to drivers´ needs.
The experts at the ERL are working intensively to evolve the vehicle into an intelligent assistant — for example, by using a combination of voice, gesture and manual controls that replicate the way humans interact. They also look at how an individual customer profi e in a digital eco system can be used, for example, to unlock a vehicle or make it available to close friends within a personal social network.
The focus here is on delivering a reliable, safe and entertaining experience at all times. Adrienne Othon has brought this insight with her after spending over 20 years in the movie industry and as a games developer. “Entertainment and safety are by no means mutually exclusive,” she says. “Why not tap the benefi ts of augmented reality and use software to brighten the view through the windshield on a cloudy day so that it looks as if the sun is shining — just like they do on a fi m set. It lifts the mood and enhances driving safety.”
Established in 1998, the Electronics Research Lab has evolved into a respected Silicon Valley institution. In addition to intelligent control systems, another large department is dedicated to researching and developing electric mobility.