The challenges of making autonomous cars a reality have proved more stubborn than expected. As a result, the widespread excitement about the technology of a few years ago has been replaced by a more cautious and sober outlook.
Perhaps this was always the likely outcome. The sheer complexity of the technology, and the difficulty of reducing the huge uncertainty in all that sensor data, means that fully autonomous driving remains a genuinely hard problem to solve.
Right now, the technology has not gone much beyond Level 2 on the five-point autonomy scale, although we have seen some limited implementations of Level 3 (highway pilot with no driver monitoring).
Level 5 (full autonomy) is not currently a viable proposition.
Is this the end for autonomous cars? Not at all. The shift will simply take place in a more evolutionary way.
For example, autonomy will initially take hold in more tightly defined situations, such as logistics or valet parking.
But even these narrower use cases could still represent significant disruption for incumbent automakers, who are at risk if they ignore the incremental developments taking place. If they do, they will miss out on a key monetization opportunity in the short term – and they could be overtaken in the long-term race to Level 5.
How should automakers be approaching this next phase of autonomous driving evolution? There are five key measures that they should be considering.
- Invest carefully in Level 2 features. The slower-than-expected evolution of Level 4 and Level 5 autonomy means the technology poses little immediate threat to automakers’ existing business models. Instead, Level 2 autonomy is likely to dominate the market in the near future. Incumbent automakers should therefore look to get on the front foot, investing early but carefully in monetizable Level 2 features. This will create a solid foundation for a future ramp-up of Level 3 autonomy and above.
- Continue driving toward software-defined automotive. Autonomy isn’t only about having the right artificial intelligence and advanced data management. It also needs the whole vehicle to be software-defined. New automotive research should therefore be focused more on future embedded software stacks, operating systems, developer ecosystems, cloud platforms, and edge technologies -- and less on the internal combustion technologies that will soon be obsolete. The whole emphasis needs to shift from treating a car as a stand-alone product to seeing it as a node within a wider connected network.
- Leverage the incumbent’s inherent advantage. Incumbent automakers have a key advantage over smaller startups: the size of their fleets. Yes, Tesla sells an impressive 500,000 vehicles annually. But the major automotive players sell 20 times that number every year. If they can connect these vehicles and start capturing the huge amounts of data they generate, they will have a ready source of Big Data on a scale other players simply can’t match. And when it comes to developing autonomous driving models and features, the more data you have, the greater your potential competitive advantage.
- Start defining the autonomous market. The scale that incumbent automakers have built up over the years also means they can be proactive about developing and defining the autonomous driving market on their terms. A reactive wait-and-see approach merely risks leaving that market open to new entrants and startups. As part of this, automakers should be looking to equip new vehicles with autonomy-related technologies even if not ultimately used by the owner. This will enable data to be gathered, analyzed, and monetized independently from the vehicle hardware.
- Partner to move faster. Autonomy is, for obvious reasons, a complex undertaking. Automakers should therefore consider bringing in external expertise – whether that’s technology companies, suppliers, or service providers – that can both accelerate time to market and improve the performance of customer-facing autonomous driving technologies. This kind of partnership approach can also bring significant savings in development costs.
