Cars go digital

Meet Veronica. If you don't already know her, you will—soon. By the end of the current model year, Veronica will be talking to drivers of about million GM cars with OnStar's Virtual Advisor, giving them stock quotes, news headlines, sports scores, and e-mail messages.

Sometime in the next few years, Veronica will turn the automotive world on its ear. She'll invade millions of cars. She'll answer questions. She'll navigate trips.

With Veronica, automakers will be able to blend complex audio and video, cellular phones, word processing, GPS data, digital radio, and much more.

And that, of course, means that cars will be more complex than ever. They'll have billions of bits of digital data whizzing through their wiring. And while all of that may be a godsend for consumers, it creates problems of maddening complexity for engineers. Among those problems:

Packaging : How do they pack a PC, radio, audio system, video unit, cellular phone, and GPS into an already crowded dashboard?

Bandwidth : With enormous video images and streaming cellular phone data buzzing through the wires, how can anyone create a system capable of handling it all? And how can they do it without doubling or tripling the size of thick wiring bundles already present in vehicles?

Safety: As engineers incorporate more systems for work and entertainment, they must ensure the signals don't conflict with data from more important systems. When vehicles collide, for example, processors have to activate the airbags. And those processors shouldn't be clogged with images of Dumbo from the video system. "We don't want to get in a situation where the navigation system interferes with the operation of the brakes," says Scott Andrews, a former Toyota engineer who is co-founder of an industry organization helping to develop standards to deal with such problems.

Veronica's role. Still, without Veronica, engineers couldn't incorporate all of those technologies into a single vehicle. The reason: The interfaces would have been too complex.

For years, automotive audio systems grew in complexity, until dashboards began to look like the control panels of a Boeing 747. Buttons were already too small and too many. And the systems were too difficult to program. As a result, some car owners never even bothered to program their vehicle clocks. Worse, engineers knew that additional complexity would further distract drivers, which could lead to accidents.

Cadillac's Infotainment system uses a simplified user interface that relies on voice activation.

With Veronica, engineers could simplify user interfaces. Why? Because Veronica changed the relationship between the driver and the electronics. Instead of leaning over, searching for a button and pressing it, drivers could now talk to the dashboard. "When the Web first emerged, people talked about GUIs (Graphical User Interfaces)," says Mike Peterson, director of the Virtual Advisor program at OnStar. "Now, as voice recognition emerges in automobiles, everyone's talking about VUIs (Voice User Interfaces)."

In the OnStar Virtual Advisor, Veronica's virtual voice is located at a remote Java-based server. In that sense, it differs from competing systems, which are using, or plan to use, on-board intelligence.

OnStar engineers say that their system offers a key advantage: Because it employs a cellular link to a remote server, it incorporates far more processing power than any on-board system could ever hope to have. OnStar's remote servers use 700 Mbytes of memory and a 1 GHz processor. Result: a vocabulary of several thousand words. Up to now, most successful voice-activation systems have used a max of 200 words—with 30 words the norm.

In its hardware approach, the Virtual Advisor contrasts sharply with Cadillac's Infotainment system (see sidebar), as well as systems developed by Ford Visteon. The Infotainment system employs all of its memory and processing capability on board (Cadillac won't give specific performance figures). Similarly, Ford's ICES (Information, Communication, Safety ,and Security) uses a 166 MHz Intel Pentium processor and a Windows CE operating system. Ford Visteon said it used significant on-board processing power because it serves as a foundation for the future. "We essentially have a desktop in the car," says Lori Markatos, product team leader for ICES.

Whichever technique automakers use, however, the underlying theme is still the same: They want to abide by a philosophy of "eyes on the road, hands on the wheel," as a way of maximizing safety for users. And the end result is that they can add features they couldn't have dreamed of just a few years ago. "Ease-of-use is going to be the key to success for automotive e-mail," notes Adam J. Weiner, a senior analyst for Gomez Advisors (Lincoln, MA), an e-commerce research and consulting firm. "It's got to be intuitive to appeal to consumers."

That's why GM and Ford both plan to continue intense development of VUIs. OnStar recently announced partnerships with General Magic Inc. (Sunnyvale, CA) and Nuance Communications (Menlo Park, CA) to enhance their existing voice recognition technologies. Ford has forged a partnership with Lernout & Hauspie (Burlington, MA) for the same reason.

Electrical challenges. Making the interface appeal to consumers, however, is only half the battle. Now, engineers face myriad problems as they deal with the onslaught of new data.

The most basic of those problems is the wiring. No longer can engineers use the old method of discretely wiring each system point-to-point. Doing so would add weight and unbelievable complexity to the process of wiring each vehicle.

What's more, most engineers already believe discrete wiring has reached its limit. The proliferation of electric motors for door locks, sun roofs, windows, and seats has added so much wiring to vehicles that there is little room left for more.

As a result, engineers know they have to use network buses—thick, two-wire cables. Up to now, the most popular of those solutions has been the CAN (Controller Area Network) bus originally developed by Robert Bosch GmbH. Many automakers are pushing for standardization of IDB-C (Intelligent transportation systems Data Bus-CAN), a version of CAN that has been used by virtually every automaker in the world. "Everyone is comfortable with the CAN bus," notes Andre Oberschachtsiek, manager of the electronics research laboratory for Volkswagen of America (Sunnyvale, CA). "The interface is simple and there's no concern over royalties."

Most automakers propose to use IDB-C for automotive multimedia applications in the near future. Along with it, they plan to employ a gateway—an application specific integrated circuit (ASIC)—that would prevent multimedia messages from clogging up safety critical applications, such as engines, transmissions, brakes, and airbags.

In the long run, however, automakers know that IDB-C will not be sufficient for multimedia applications, which need greater bandwidth to be successful. That's why many car manufacturers are now considering MOST (Media-Oriented Systems Transport), a fiber optic bus. MOST offers speeds approaching 25 Mbits/sec, compared to about 250 Kbits/sec for the copper-based CAN buses. MOST, however, has issues that still need to be addressed, engineers say. Its interfaces are more complex than those of CAN buses and automakers say they are unsure about its royalty issues. Companies behind the development of MOST could be seeking as much as 30 cents per connection, which could be costly when multiplied over hundreds of thousands of vehicles and millions of connections.

To save on potential costs, European automakers are also considering application of still another system—the LIN (Local Interconnect Network) bus. LIN, which is supported by a group including BMW, DaimlerChrysler, Volvo, Volkswagen, Motorola, and Volcano Communications, is de- signed for simple on-off devices. Its bread-and-butter applications include the electric motors used in car seats, door locks, windows, sun roofs, HVAC flaps, windshield wipers and a host of other simple applications. Proponents of LIN say it could accomplish most of the same tasks as the SAE J-1850 bus, an automotive bus that never properly standardized, but at a lower cost than other buses. LIN's tradeoff for that lower cost is speed: Operating at about 20 Kbits/sec, it is about one-twelfth as fast as CAN, and one-six-hundredth as fast as MOST. "LIN is not going to replace CAN buses," says Oberschachtsiek. "It's for applications where it doesn't matter if the information comes 50 milliseconds sooner or later."

Standards needed. Even if automakers make liberal use of new bus technologies, however, vehicle chief engineers still face a daunting task: which electronic features to choose, and which not to choose. That's particularly difficult when considering the rate of change in the electronics industry, and then coupling that with the two to three years that it takes to design and build a new vehicle. Even the fastest automakers—the ones talking about 20-month cycle times, can't keep up with the frantic pace defined by Moore's Law on the doubling of processing power every 18 months.

Automakers agree that the answer to that problem is standardization of electrical architectures. Standardization would enable them to plug new electronics into vehicles at the last moment before production. Standardization, however, is far easier said than done. "The reason we've seen such strides in the PC industry is that everyone is developing their products against the same platform," says Andrews, who was instrumental in the formation of the Automotive Multimedia Consortium (AMIC), a group dedicated to standardizing the bus architectures for multimedia applications. "The auto industry is almost the exact opposite."

The AMIC consortium, which includes 12 of the world's biggest automakers, hopes to settle on two standard network buses—a low-speed bus and high-speed bus. Such standardization would not only enable automakers to reduce thickness of wiring bundles; it would eliminate their need to re-engineer vendor products. Many automotive engineers now complain that they spend months re-engineering because vendor products so seldom comply with the automakers' proprietary vehicle architectures. Toyota, for example, has said that it uses four different vendor-supplied navigation systems in its vehicles, then re-engineers each to fit its own electrical architectures. "All of the suppliers use different microprocessors, interfaces and software," says Andrews. As a result, automakers must often "go back in and tear up their code," he says.

As automotive multimedia competition heats up, however, OEMs will have to agree on standards, experts say. "This is a group that will fight over the placement of an extra screw in a vehicle," notes George Fry, president of Aviso Micro Technology (Glendale, AZ), an automotive consulting firm. "But they're going to have to agree on some common bus structure within a year. It's in their best interest to do it."

If they do, consumers will benefit. Industry experts expect other automakers to follow Cadillac's lead and, using variations on Veronica, bring e-mail to their automobiles. "People are getting addicted to their information sources," says David Cooperstein, director of online retail for Forrester Research (Cambridge, MA). "Once they've had a taste of these capabilities, they won't want to give them up."