Autonomous Cars Need Million-Mile Reliability

In the U.S. today the average car is 11.5 years old with about 150,000 miles (241,401 km) on the odometer.

That means they’ve reached their traditional life expectancy. But as automakers turn their attention to ride sharing and car sharing, they’re going to have to design cars to last longer than that – a lot longer.

With regular maintenance and a lot of tender loving care, cars can last a long time. Taxis in big cities that are driven two shifts a day typically rack up 70,000 miles (113,000 km) to 100,000 miles (161,000 km) a year. But ride sharing and car sharing could make those numbers look puny.

These are the findings that came out of a presentation titled “The Impact of Autonomous Vehicle Technology and Ride Sharing on Design for Reliability” by James Wasiloff, an engineer at BAE Systems, a British multinational defense, security and aerospace company.

His premise is that automakers are going to have to revamp the duty cycles and the amount of testing they do. Instead of designing cars to last 150,000 miles as they do today, he says the vehicle’s design life will have to jump to 1 million miles (1,609,344 km).

It all has to do with simple mileage math. The average car owner today uses his or her car about one hour a day and lets it sit for 23 hours. At this rate, the average owner drives about 13,000 miles (21,000 km) a year.

But with ride sharing, the situation will flip. Wasiloff projects that a car used 22 hours a day would rack up at least 233,000 miles (375,000 km) a year in city driving and much more in highway driving. In three years’ time that car will have racked up at least 700,000 miles (1,127,000 km), but no car and none of its components are designed to last that long. At this rate the powertrain alone would have to be replaced every six to 12 months.

With autonomous cars, the duty cycle and testing becomes even more critical. Wasiloff points out that LIDAR, radar, video cameras and inertial measurement systems will need to be validated for much longer lifecycles than they are today.

He says they need to be tested in severe snow and ice conditions and for long-term solar radiation degradation. The interfacing of the components, the wear and tear and the manufacturing tolerances all will need to be tested for longer life.

Moreover, automakers will need to add redundancy to their autonomous cars so that the failure of any of these components or systems does not affect the others. And that’s going to take more testing.

Fortunately, automakers are going to have new tools at their disposal to make their cars more reliable. The more connected cars become, the more data there is to monitor them. Automakers will be able to collect that data for reliability prognostics, similar to what the manufacturing people call predictive maintenance.

Using this information and big data analytics, automakers will be able to identify any degradation in these systems before they become a problem and take corrective action. Like they saying goes, it’s far more effective to fix something before it breaks rather than waiting for it to break.

I’m fascinated by these findings. While others have talked about the need to drastically boost the Mean Time Between Failure of autonomous cars, if only from a safety and liability standpoint, Wasiloff points out that this mean time will need to stretch out over many more miles than today’s cars.

Autonomous cars point to a promising future, but a report like this shows there are a lot of tedious details to work out before we can safely put them in the hands of the public. And it all has to start with changing the duty cycle.