In the coming years, electronics is destined to play the strategic role in about every aspect of automotive technology and innovation. By the end of this decade, electronic systems will supervise and control such mission-critical functions as braking and steering. In an even shorter time, automobile infotainment systems will become sophisticated enough and versatile enough to rival in-home entertainment in terms of the user experience and will be superior to home systems in ease of use.
As streaming multimedia makes its way out of the home and into the car, consumer buying decisions will be increasingly based on features delivered by electronic systems and the integrated circuits that power them.
So it should surprise no one that auto-industry pundits predict 90 percent of tomorrow's innovation will be based on electronics, including infotainment, advanced driver assistance (based on radar and vision), and X-by-wire–engineers' shorthand for brake-by-wire, drive-by-wire, park-by-wire and other basic and critical control and safety functions.
At the same time, the conventional electronic-system architectures that have slowly evolved–pretty much as handmaidens to mechanical systems–are approaching an evolutionary dead end. In the not-too-distant future, extensions of existing standards, engineering practices and architectures will be incapable of satisfying the automobile's growing appetite for electronic innovation.
The dominant way to distribute control information around the car from one subsystem to another is more than a decade old. The so-called CAN (for controller-area network) bus has limited capacity for moving data; it also has the electronic equivalent of attention deficit disorder.
At some point, CAN's data bandwidth of 1 Mbit/second will no longer be able to handle all of the communications needed for controlling the car. Streaming audio and video, of course, need orders of magnitude more bandwidth.
CAN, as conceived, was never intended to make decisions within a predetermined–and very short–period of time. But information critical to actuating a car's brakes requires exactly that: response in a few milliseconds.
Today's cars handle CAN's lack of “determinism” with dedicated communications links between every actuator (a brake pedal, for instance) and the implementing system. There is no delay, because multiple systems are not trying to use the bus at the same time, as occurs in typical data networks.
Obviously, CAN's point-to-point architecture works extremely well. Drives can have 100 percent confidence that when they apply the brakes, the car stops.
Multiple, independent systems create unnecessary complexity and add significant cost. Today's high-end auto has up to 100 electronic control units (ECUs), each with its own microcontroller chip and other components. Because each operates pretty much independently–and many must be wired back to the driver's control panel–the wiring harness is becoming as large as a boa constrictor. But the processing for many functions now being done by CAN could be handled by just a few high-performance microprocessor chips.
Engineers agree that the time to integrate multiple functions into fewer ECUs has come: This is all about simplicity. But it requires more-open communication channels between actuators and the systems they implement. And that requires that the in-vehicle network be deterministic–that is, capable of guaranteeing a response within a specific time period.
The key players may not yet be known with certainty, but one thing is clear: In the next decade, driving a car will be profoundly different from what it is today, shifting from the act of “driving” to the experience of “enjoying.”