Selective Redundancy In Cars

The automotive industry has been fish-tailing its way through design strategies and electronics architectures, but it finally appears to be honing in on a strategy that actually might work. This doesn’t mean fully autonomous vehicles will take over the road anytime soon, but at least it points carmakers in the right direction.

The auto industry has been in panic mode ever since Tesla, Waymo, and a few others began proving to the world that cars really can drive themselves. Unfortunately for most of us, the hype preceded reality. It may be decades before a vehicle can safely weave in and out of city traffic smoothly and without a steering wheel. But in the span of the last five years, the auto industry has transitioned from metal-bending to electrical engineering. And for an industry that has been around for nearly 140 years, this is a radical shift.

In the first few years of this transition, carmakers swerved from one idea to another. They began with the idea that sensors would be able to collect data and send it to the cloud using 5G millimeter wave technology. That proved to be thoroughly misguided, based on misconceptions about the time it takes data to travel to and from a car, as well as the reliability of 5G.

So instead they began building data centers that would sit in the trunk of a car. That was proven to work, but the price tag for a system capable of navigating all possible corner cases was somewhere in the range of seven figures per vehicle. Even with economies of scale, that was untenable.

Rather than scrap that idea immediately, they decided to trim costs elsewhere by doubling up on fail-over systems. For instance, if the engine control system failed, why not have the infotainment system or some other circuitry in the car take over? There were a couple of good reasons. First, the infotainment system is not ASIL-D qualified. And second, any system that is in constant communication with the rest of the world is highly susceptible to cyber attacks, which in the case of cars also makes them unsafe.

Finally, after several years of wrong turns, carmakers recognized that the way forward was to move more intelligence out to the sensors, rather than processing centrally, and to sharply limit the data flow inside a vehicle. That led to yet another way to cut costs without sacrificing reliability, and that’s where we are today. The current thinking is that some redundancy is necessary even though it costs more money up front, but that not everything needs to be redundant.

Carmakers initially panned the idea of any redundancy because it was too expensive and added too much weight to a vehicle, and with everything centralized it would have required another supercomputer. But with only some functions being redundant, this begins to look like a strategy that actually makes sense.

It’s not necessary to replicate everything, but if the motor control system or braking system fails, then people can get hurt. So there are places where it makes sense to spend extra money — something the auto industry has been loath to do in the past, as evidenced by the fact that most cars no longer have a real spare tire anymore — and places where it clearly does not. But in an increasingly autonomous electric vehicle, those places are different than in a vehicle controlled by a human driver. In fact, some of that redundancy cost may be a digital twin that isn’t even in the vehicle.

The auto industry has spent a half decade in a period of turbulence, but it finally appears to be heading in the right direction. From here on, progress likely will proceed much faster.