Paving The Way To Autonomous Driving

Over the past couple weeks, four major carmakers began pairing off to jointly develop autonomous vehicles. Numerous reports say Ford will sign a deal with Volkswagen, and BMW is working on Level 4 self-driving vehicles with Daimler, the parent of Mercedes Benz.

While this speaks volumes about the enormous cost of developing artificial intelligence systems to drive vehicles, it also points the way to more consistency in vehicle behavior. The automotive industry historically has been very effective at squeezing every last fraction of a penny out of the supply chain, and this plays to the industry’s roots.

But there’s much more to this deal than just counting pennies. For one thing, there is a need for far more data than any company can develop on its own. While companies report millions of miles driven, there is a need for billions of road tested miles. Simulation is a critical component, but simulation plus real world data to map corner cases is equally critical. Outlier data can then be integrated into models, which in turn can be regionalized so that all vehicles in Sweden can behave predictably on icy and wet roads, while those in Saudi Arabia can be optimized to deal with sandstorms.

These kinds of deals also start to add consistency to data across carmakers. This is critical for vehicle-to-vehicle and vehicle-to-infrastructure communication, but it’s also critical that cars behave the same way to avoid accidents rather than hitting each other while trying to avoid objects or dealing with spin-outs on a snow-covered road because they are not in sync.

While getting autonomous vehicles on the road faster is everyone’s goal, getting them to behave well around each other is equally important. That requires them to process data the same way, using the same algorithms and the same inference models. This is not where carmakers should be competitive, and these deals are an indication they are coming to that realization.

It’s also a recognition that putting all of that control into the hands of a standards body is far too slow and inefficient. As more electronics are added into vehicles, the speed at which the technology changes will accelerate. This is far different from mechanical designs, where standards can be set well in advance. As electronic content grows, standards need to be generated from the technology and not the other way around. Moreover, those standards need to utilize real-world data involving all brands of autonomous vehicles that will share the public roadways.

Carmakers are taking concrete steps, whether it’s due to the impact of electronics or simply the exploding cost of developing AI-based systems, to make autonomous driving a reality. It’s better to team up and get to market with a competitor or two than to not get there at all. That also helps with liability issues, where it’s easier to justify best practices across an industry than a single carmaker. And it doesn’t hurt when autonomous vehicles all move at the same speeds and react in identical and predictable ways, which can only happen when data is shared across the industry.

But what will differentiate all of these vendors besides the price when cars really can drive themselves? So far that isn’t clear, and that has to be worrying a lot of automotive companies with a long history of making some very fine automobiles. Consumers may not care if they can hear the clock or the rev of the engine if everyone is moving at exactly the same speed.