Two notable milestones on the road towards autonomous driving have been passed this week, by Hyundai in South Korea and Yandex on the snowy streets of Moscow. Hyundai’s was significant for being the first major show of SAE Level 4 autonomous driving in an electric car powered by hydrogen fuel cells, which promise to solve the range problem for electric cars but still have work to do in reducing size and weight. This involved five cars driving the 190km from Seoul to Pyeongchang, site of the Winter Olympic Games, at speeds up to around 110kmph.
Hyundai claimed this demonstrated for the first time some of its most recently developed autonomous technologies, including the ability to pinpoint the vehicle’s location precisely using external sensors in the event of GPS signal loss, such as in tunnels, as well as being able to negotiate constricted areas such as toll gates safely.
Three of the vehicles are based on the manufacturer’s NEXO SUV, due for launch in South Korea in March, although not at this stage cleared for all the Level 4 autonomous driving features, which are scheduled for 2021.
SAE Level 4 is the current gold standard for autonomous driving, where the car can drive itself nearly all the time except in severe weather or unmapped areas, requiring drivers to be ready to take over the wheel but able to engage in other activities, or even nod off to sleep.
Most manufacturers, Hyundai included, intend to jump straight to Level 4 from Level 2 which is where many Tesla vehicles currently are for example, requiring the driver to remain alert at the wheel with just some cruise control and basic automation. Level 3 is seen as dangerous middle ground where the car takes over a lot of the driving but requires the driver to be fully on hand to take over some safety critical functions at short notice.
Elsewhere, the Yandex demonstration was interesting in that while it lags behind Hyundai’s capabilities in most respects, it is reaching towards the final Level 5, with its handling of roads covered in ice or snow. This is normally deemed to be beyond Level 4, but obviously in regions with a lot of adverse winter weather, it cannot be postponed that long if autonomous driving is to be of any value.
Such conditions are challenging, not just through lack of wheel grip but also because road markings normally used by the car’s sensors are obscured. Optical vision systems are then less effective and so cars have to rely less on lidar (Light Detection and Ranging) for assessing distances and more on radar, because that can take measurements from the wider landscape. Optical vision can still play a useful role but requires cameras to be located differently, focusing more on rear and side views, while integrating with radar.
So Yandex seems ahead on winter driving but behind on coping with the hazards of roads under normal conditions, such as pedestrians and heavy urban traffic. Russia still prohibits self-driving and Yandex has yet to test its cars on public roads and so is unlikely to bring out Level 4 cars for quite a few years.
In any case, being Russia’s equivalent to Google (but distinctly smaller), it relies heavily on partners, notably the country’s leading truck maker Kamaz to develop the self-driving vehicles themselves. These two in turn have been collaborating with Daimler and the Russian government-backed research organization NAMI, to develop an autonomous minibus shuttle that could become part of a shared mobility infrastructure.
So while Yandex itself has developed the computer vision, radar and recognition technology, NAMI has provided testing facilities and Kamaz the manufacturing.
On the other hand, Hyundai, as one of the world’s big five automobile makers (the others being Ford, GM, Volkswagen and Toyota), has great resources and has developed a lot of the self-driving specific technology internally, even if it relies on usual suspects for commodity components. It has developed its own fuel cells because it regards this as a competitive differentiator by being able to reduce size and increase range without sacrificing performance.
Hyundai does though have one key partner for autonomous driving – the US start-up Aurora Innovation, one of whose three founders was Chris Urmson, former Google head of self-driving.
Aurora is focusing on the autonomous driving technology stack, embracing sensors, software and data services, with Hyundai unlikely to rely on this for more advanced functions designed to stand out from the competition. That is because Aurora is also partnering with Volkswagen and so would be unlikely to gain significant competitive advantage that way, unless there are some specific projects within the collaboration.
Hyundai has already emphasized that autonomous driving will be of limited value without all the associated services around navigation, infotainment and personalization. At the same time autonomous, driving does require high networking bandwidth to operate efficiently and safely under all conditions, with the leap from Level 4 to Level 5 likely to be at least as great in technical terms as the others put together. So as part of its South Korea demonstration, Hyundai was talking up the value of its in-car 5G connectivity, even though this will be of limited use in most parts of the world for some years to come, until the cellular infrastructure has been upgraded accordingly.
However, 4G or even 3G will be sufficient for some of the novel features Hyundai has been demonstrating, such as “Home Connect,” a car-to-home technology enabling passengers to access and control IoT devices installed in their smart home. They can view home camera images in real-time, control lighting, remotely lock doors and manage home energy systems. Users can also manipulate relevant car systems remotely via a smartphone and Hyundai has been working with Google as well to add voice to these capabilities.
It will be the performance of core autonomous functions during trials and limited deployments that will determine how quickly the full Level 5 deal, or for that matter Level 4, becomes a reality on a large scale. Inevitably the bar will be set higher for autonomous than human drivers and just a few deaths clearly attributable to system rather than human failure, or even through misunderstandings over which is in control, will hold back the necessary regulatory relaxation.