PARK RIDGE, Ill. The auto industry's search for a safety-critical data bus intensified this week as two major automakers swung their weight behind different systems.
In a move that could have a powerful impact on the industry's decision to back a single standard, General Motors Corp. announced at the VDI Conference 2001 in Baden-Baden, Germany, that it is lending its support to the FlexRay Consortium, which already includes DaimlerChrysler, BMW, Motorola, Philips Semiconductors and the Bosch Automotive Group.
Days earlier, PSA Peugeot Citroen lined up behind the Time-Triggered Architecture Group, which includes Audi, Volkswagen, Honeywell and Delphi Automotive Systems.
The announcements added more uncertainty to a situation that was already clouded by claims, counterclaims and talk of possible legal action. Each group has said that its technology is best-suited to handle the safety needs of future automobiles, particularly as they begin to incorporate such features as drive-by-wire, adaptive cruise control and collision avoidance.
Automotive analysts said, however, that the addition of GM could give the FlexRay Consortium more leverage in its effort to serve as the basis for an industry standard. "GM's support will go a long way," said Paul Hansen, publisher of The Hansen Report on Automotive Electronics. "GM and DaimlerChrysler make up a large percentage of the world's vehicles, and that's going to be hard to ignore."
Safety vs. flexibility
Still, companies on both sides said that the choice of a safety-critical data bus standard is far from settled. The debate over the best technology will continue for some time, analysts said, in part because the industry sees it as such a key element of future vehicle design.
Automakers have been pushing harder for the development of a safety-critical bus for the past two years because of the looming prospect of drive-by-wire systems, which include steer-by-wire, brake-by-wire and throttle-by-wire. Such systems let automakers eliminate heavy hydraulic actuators, and they offer the possibility of creating smarter, more efficient components that can be connected to a network bus. Automakers and vendors know, however, that a reliable, fault-tolerant bus is needed for such applications. Controller-area network (CAN) buses, which are commonly used for powertrain and other automotive controls, are not considered reliable enough for drive-by-wire.
"The issue with CAN is that it is only event-based, so there is always the possibility that a message won't get through," said Hansen. "For important applications, you need time-triggered architectures, because as time rolls along, you always have a slot for important messages."
The time-triggered architectures being used by the TTA Group and the FlexRay Consortium are fault-tolerant, enabling the electrical bus and controllers to continue operating in the presence of a failed sensor, short circuit, transient software glitch or other problem. As a result, safety-critical systems remain up and running.
Still, the two groups are not in agreement on how best to implement such fault-tolerant buses. The TTA Group employs a Time-Triggered Protocol originally developed at the Technical University of Vienna and licensed by TTTech Computertechnik GmbH (Vienna, Austria). TTP, which is said to have robust protection against signaling errors, uses a time-slot-based (time-division multiple access, or TDMA) architecture. Slots at all nodes are controlled by a distributed correction scheme, and the architecture includes a "bus guardian" that prevents a single node from grabbing control of the bus because of software error. Under this scheme, if controllers don't receive signals at prescribed times, they can determine on their own if there's a problem.
Proponents of the TTP system say that it provides the safest solution. "During the last months, we came to the conclusion that TTP is a proven technology . . . able to address all our needs in the field of reliable communications," said a PSA Peugeot Citroen representative in a statement issued after the company opted to employ TTP protocol in its by-wire prototype (but not necessarily in production vehicles).
PSA Peugeot Citroen's sentiments matched those of Audi engineers, who reiterated their support for TTP this week. "Our point of view is still 'security first,' " said Walter Streit, senior engineer responsible for bus systems and air bags at Audi AG. "And we see TTP as being better in terms of security."
That point of view was also echoed by a recently published paper, "A Comparison of Bus Architectures for Safety-Critical Embedded Systems," based on research supported by NASA Langley Research Center. The paper considered TTP to be more mature, while FlexRay is "still under development."
Based on recent developments, however, proponents of the FlexRay system believe that it offers the same security as TTP, plus the flexibility to serve in an automotive production environment.
Based in part on the Byteflight system, FlexRay owes its determinism to an architecture that uses flexible time-division media access time slots for data. The system targets a maximum of 64 nodes and offers two channels of 10-Mbit/second data rates. It also employs a bus guardian, which eliminates the so-called "babbling idiot mode," in which a faulty sensor transmits a stream of meaningless data and locks up the controller.
Engineers from Motorola and General Motors said this week that the key difference between FlexRay and TTP is FlexRay's use of both static and dynamic components. FlexRay's static portion is time-triggered and synchronized to a clock mechanism. Its dynamic portion, however is event-driven.
"With FlexRay, you can choose to have a static and a dynamic part," said Florian Bogenberger, a system research and development engineer with Motorola in Munich, Germany. "And you can make the dynamic part as large as you want, or make the static part as large as you want."
That flexibility is said to be key to enabling easier manufacturability for automakers. "We need the flexibility so we can accommodate our mix of models," said Ben Baker, director of the Electrical Product and Process Center for General Motors (Warren, Mich.).
Baker said that GM typically employs electronic features across several car models, then looks to change those features from year to year. Such feature changes can become maddeningly complex if the electrical architecture needs to reengineered to accommodate them, he said. And GM engineers fear that, with TTP, that might be the case.
"The big problem is verification and validation," Baker said. "After you put the system back together again, you've got to make sure that the systems still meet their requirements. We've got 20, 30, or even 40 'micros' talking to each other, and we've got to make sure they get the right data at the right time." By flexibly altering FlexRay for those needs at the outset, Baker said GM doesn't expect to do a lot of re-engineering from year to year.
Motorola engineers said that such OEM interest was the key to their interest in FlexRay. "Before we invest in a protocol, we want to make sure an OEM is going to back it," said Jim Trent, transportation business operations manager in Motorola's Semiconductor Products Sector (Austin, Texas).
Automotive industry analysts point out that FlexRay now has the lion's share of vehicles in its corner. In 1999, GM, DaimlerChrysler and BMW together produced more than 14 million vehicles worldwide, while PSA Peugeot Citroen, Volkswagen and Audi accounted for about half of that, Hansen said.
Call for unity
Despite the differences, engineers from both camps said this week that they hope to pull together to create a single standard. "We want the industry to come to a single standard, and we want it to be open," said Georg Kopetz, a director of TTTech. "If there are any requirements not met by TTP, we can, together with the other players, alter the standard."
TTP executives warned, however, that they will look carefully at the FlexRay protocol to ensure that its static portion does not in any way build on TTTech's patented concepts. TTTech, which developed its time-triggered concepts before the formation of the FlexRay Consortium, initially worked with DaimlerChrysler and BMW.
"We don't want the impression that TTP is involved in FlexRay," Kopetz said. "There is no contract enforced between the TTP patent-holding company and the FlexRay Consortium."
Ultimately, analysts believe that the two groups will be forced to agree on a single standard, if for no other reason than the good of the industry. "If there's not much difference between the protocols, then it would be ridiculous to have two of them," said Hansen of The Hansen Report. "In the end, the decisions may turn out to be more a matter of politics than technology."