Autonomous transport - help the driver before cars help themselves

Chris Middleton Profile picture for user cmiddleton May 18, 2020
Summary:
A more moderate and realistic vision of the driverless future than we are used to, from Dell’s CTO of Automotive and AI.

cars
(Pixabay)

The mass arrival of driverless, autonomous cars is at least 10 years away, and they will need a dedicated infrastructure – both built and technological – to support them in our cities. Rather than mingling with other traffic, driverless vehicles will need their own separate lanes – like busses, cycles, and taxis – and to be kept away from pedestrians.

That’s a very different vision to the one we are normally presented with of on-demand autonomous transport; most technologists have imagined these vehicles blending seamlessly with urban life. But that’s the view of Dr Florian Baumann, CTO EMEA of Dell Technologies’ Automotive & AI business, who explains:

I'm very conservative. I do not see that self-driving cars will become a reality during the next 10 years. Companies will start with use cases like valet parking, or self-driving in limited areas, such as between Los Angeles and Las Vegas. It will be at least 10 years or more before we reach Level 5 autonomy at scale.

Gear shifting

Dell is a name more usually associated with beige boxes than with an autonomous transport market that is dominated by the likes of Waymo (Alphabet), Uber, Tesla, Baidu, GM, Ford, Jaguar LandRover, Toyota, and BMW.

The opportunity for Dell – which launched its own IoT division in October 2017 – lies in the supporting data and computing infrastructure needed for Connected Transport services, both in the cloud and in the critical edge environment. It believes that employing a data-first strategy will be key to success in this space.

According to the company, new data-driven revenue streams are forecast to grow to $151 billion a year worldwide by 2035, a massive 7,500% increase on current figures. But before the global market gets to that point, technologists’ and policymakers’ focus should be on developing Advanced Driver Assistance Systems (ADAS). Baumann believes these are an important milestone on the journey to fully autonomous vehicles.

Level 1 and Level 2 ADAs are already available in many new vehicles, but by 2030, analysts  at Frost & Sullivan predict that one in four cars sold globally will have at least Level 3 autonomous driving capabilities. Credible?  Baumann says:

ADAS functionalities up to up to Level 3 – lane change departure warnings, autonomous braking, driver drowsiness detection, driver distraction detection, in-car driver monitoring systems, and so on – all these functionalities will become reality in low-cost vehicles.

The gradual acceptance of these systems will be vital to challenging the view that driving is what humans do, not machines – something the US Department of Transport began to do in 2018 when it extended the legal definition of ‘driver’ to include autonomous systems:

It’s important to get the acceptance of the driver. There was a study a couple of years ago which found that 80% of motorists are deactivating driver assistance functionalities inside of their vehicles. So these technologies are not yet accepted. This is, from my perspective, is definitely a waypoint on the journey to autonomous driving.

The flip side of the coin is those drivers who place too much trust too quickly in assistance technologies, perhaps believing that systems such as Tesla’s Autopilot live up to their misleading names. In this way, unwary owners may be abdicating responsibility for their own safety.

That seems to have been a factor in the 2018 Tesla crash in which a driver lost his life while the car was running under software control – one of a number of accidents in that year that damaged public confidence in autonomous transport.

Either way, the cultural obstacles to driverless cars remain legion, especially in the US, where driving remains the most common job in many states. Some US cities – such as Los Angeles – lack public transport infrastructures, while the idea of the lone driver out on the freeway is at the heart of American pop culture. And post COVID-19, the appeal of public transport is likely to diminish further…

While US government statistics reveal that 94% of deaths on the road are down to driver error, impairment, or distraction, anyone who believes that a century of American auto culture is going to be swept away overnight by Waymo, Tesla, and Uber is kidding themselves.

In China, autonomous vehicles is likely to be more quickly accepted, as the culture of car-ownership and freedom on the roads is not so deeply ingrained there. Yet Baumann is optimistic that change will come in the Wes and that it is “just a question of time”.

However, culture isn’t the only obstacle to autonomous transport. The technology itself is an issue – and not just within the car, in terms of sensors, computer vision, or AI systems. In a world where many citizens still struggle to get viable broadband services in their homes and others find themselves tapping impatiently at their phones in mobile ‘not-spots’, the real-world technology barriers to driverless cars should be obvious:

There are still regions where you don't even have 3G right now and so only have an edge GSM connection. I guess it may take years until we get 100% 5G coverage, so we can't rely on having a permanent 5G connection. So what you have to do is cache the data inside the vehicle, pre-process it. So, you need in-car storage and compute, or compute and storage on the edge.

Outside of network hotspots or city limits, driverless vehicles may have to rely on laser communications or links to satellites, such as SpaceX’s Starlink constellation, wherever 5G connections are patchy or non-existent on the ground, says Baumann – particularly in rural areas.

For Dell, it all comes down to data, rather than to the cars themselves. By 2030, the company estimates that 90 million connected and autonomous vehicles will each generate up to 10 Terabytes (TB) of data a day – or one Zetabyte (ZB) a day across the industry. With that much data being created and shared, it’s essential for OEMs and Tier-1 suppliers to push a data-first strategy, says Baumann:

One of the areas for us is data management at scale. These cars will generate huge amounts of data after they are prioritised and on the street, because they are continuously communicating with the cloud and with the on-premise infrastructure.

We are active in this space by delivering products that enable our customers in the automotive industry to develop ADAS functionalities and self-driving cars. This starts in the data center by delivering storage appliances, and then there’s computing at the edge, and computing and storage inside of the vehicle.

My take

A more moderate view of the future than we are used to – and perhaps a more persuasive one.

It’s clear that the road to driverless cars will be long and full of twists, turns, and unpredictable events. Our cities will be shared spaces for many years to come and once out onto country roads, it would be foolish to assume that no humans will be behind the wheel.

In the meantime, building a high-speed, resilient, secure, consistent, and reliable data infrastructure will be essential; without it, the driverless future is just a pipe dream.

Loading
A grey colored placeholder image