Brick by brick: autonomous vehicles need rock-solid foundations

In order to avoid self-driving vehicle catastrophes later, there are certain things OEMs must get right now, says Wind River. By Xavier Boucherat

History might well look back on 2016 as the year when autonomous driving moved from science fiction to inevitability – that is, in the eyes of the public. The automotive industry has been pushing the shift for several years, and some companies have been more vocal about it than others. Google’s self-driving vehicles have completed over two million miles of testing on public roads, and in 2015 Tesla’s Elon Musk went as far as to say fully autonomous driving was a ‘solved problem’.

But it’s only since the start of the year that the big players – with much more to lose should it all go wrong – have begun to publically throw the same weight behind the idea, with ambitious timelines and out-in-the-open testing. Ford, for example, announced in August that it plans to have a high-volume, SAE level 4 capable vehicle on the road by 2021, for use in a ride-hailing or ride-sharing service. In the same month, Volvo and Uber announced they were working together on a US$300m project to develop new base vehicles that can incorporate the latest advances in autonomous driving. Engineers from both the Swedish OEM and the ride-hailing service will use Volvo’s scalable platform architecture (SPA) to build a vehicle with the connectivity, safety and redundancy requirements that tomorrow’s cars will demand.

“I think, unbeknownst to much of the general population, automakers have been exploring different levels of automation for many years,” says Marques McCammon, General Manager of Connected Vehicles Solutions at Wind River, an Intel subsidiary specialising in embedded software. Some major OEMs, he points out, have programmes dating back ten to 15 years.

That said, he agrees that a combination of new entrants and government agency interest in the potential safety benefits of autonomous vehicles has accelerated the conversation.  “OEMs are now under great pressure to demonstrate leadership,” he suggests. “In my opinion, the so-called disruptors and new technology companies now entering the automotive space have prompted many different discussions, such as the idea of mobility as a service. So whilst start-ups such as Lyft and Uber are bringing new mobility models, I think there was need for the market to understand that OEMs are still very much invested in the future state of mobility, and will take a leading role.”

OEMs are now under greater pressure to demonstrate leadership. In my opinion, the so-called disruptors and new technology companies now entering the automotive space have prompted many different discussions

The challenge will be reconciling the conservative approach of the automotive industry with the quick, fluid nature of tech-savvy disruptors. This will be particularly essential as the Internet of Things takes a greater shape, and all devices posses some degree of connectivity. “I like to think of what we provide as the bridge between Silicon Valley and Motor City,” says McCammon. “In the tech world, software moves very rapidly, and enters the market in various stages of maturity. By comparison, automotive moves in a slower, more calculated fashion, with an extremely low tolerance for risk.” Automotive, he suggests, falls within a category of industry alongside, for example, medical, defence and aerospace, where reliability and safety are indispensable: “These are systems that, quite frankly, have to work when you call on them.”

Laying the foundations

To bridge this gap, McCammon explains that Wind River must provide services that can enable OEMs to make software a much bigger parts of their business. “Our focus today is on delivering a framework,” he says. “The analogy I use is like building a house. Most people, when they fall in love with a house, fall in love with its exterior aesthetic, or its kitchen, or bathrooms, but the real value of it always begins with the foundations it’s built on. A solid framework can facilitate the interactions between consumer and vehicle, between computing systems and controls, and between the vehicle and its outside environment.”

To that end, the company has developed its Helix Chassis platform, which brings together software, technologies, tools and services that it hopes will help OEMs to build into their vehicles the high-tech features demanded by tomorrow’s customers. The platform breaks down into three categories – Helix Drive, Helix Cockpit and Helix Car Sync. The first of these, Drive, is designed for the speedy but safe development of critical systems, such as autonomous driving features. These systems, explains McCammon, are fundamentally closed, and will not be receiving regular inputs from outside parties: “Consider your smartphone: you’ll download apps, get updates on a regular basis, and have data constantly moving in and out of it. It’s a social device. Drive is designated for systems that can’t operate this way; if they’re going to communicate with the overall vehicle system, it’s through very controlled channels.”

The next system is Cockpit, which is designed to provide functionality for in-car devices such as infotainment systems, head-up displays, and other places where the consumer interaction is more casual. This could allow consumers to download new apps, or record entertainment. What’s important, McCammon stresses, is that Drive and Cockpit be allowed to ‘mingle’, or meet at certain points where they can communicate. The exchange of data between the two can improve the driving experience, he explains, without compromising the security of the vehicle. “We don’t believe in a future where the in-vehicle infotainment system and automated driving are completely independent of each other,” he says.

We don’t believe in a future where the in-vehicle infotainment system and automated driving are completely independent of each other

The final piece of Helix, Car Sync, is a solution for updating vehicle modules via over-the-air (OTA) updates. Such systems need to be able to perform a variety of tasks, from updating a single line of code in one device to updating the entire vehicle, rolling back updates and integrating security systems. If the software has a cold, says McCammon, there needs to be a way to give it medicine, and when potential threats emerge, they need to be countered. This is critical, says McCammon, because compared with industries that have already embraced connectivity, automotive faces potentially far more complex challenges.

“Wind River has been involved in the aerospace industry for more than 30 years,” says McCammon, “and making these systems secure involves being able to identify threats, installing redundant paths for when systems fail, signalling a time to switch to a redundant path, and ensuring there’s no latency in these transitions. This is essential as an accident can happen within a split second. The level of security in aerospace is already mind-boggling. But when you consider automobile, you have an order of magnitude more vehicles interacting, and they’re only interacting in 2D – you lose the option of changing altitude to avoid threats and collisions. So the challenge is significant.”

All of those algorithms and all the computing in the world are critically important, but if the data gets mixed improperly, we’ll find ourselves dealing with more critical and catastrophic issues

Off the highway and onto the streets

Just how close does Wind River believe the industry is to fully realising self-driving cars? The truth, says McCammon, is that there is unlikely to be one particular moment where everything falls into place. Certain challenges have already been solved, while others remain some years off. He believes the biggest of these challenges will be designing vehicles that can handle the first and last mile. Over the next 18 to 24 months, he explains, the industry can expect to see levels of comfort with self-driving technology in highway circumstances increase. This is because highways are controlled environments, with vehicles moving in a single direction at more or less the same speed, meaning a much lower chance of on-road anomaly.

Urban driving conditions present a whole set of new challenges. “The value of human judgement is the fact you’re detecting things in your periphery with all applicable senses, and that you can make a qualitative judgement in real time,” says McCammon. The task at hand is to build devices with systems that can make the judgements whilst remaining secure and isolated from corruptive, outside threats, and once these judgements are made, to send them to the point of actuation within a vehicle whilst remaining secure.

“We see many different approaches to this question,” he says. “There are some that are completely algorithm-based, in which we programme the vehicle with the things we know and construct complex arrays of sensors and communications protocols. We also see discussions around deep-learning, and I believe this area is very promising, but challenges remain in advancing the state and strength of these algorithms to a point where the general public will feel comfortable with them.”

Fundamental to any discussion moving forward will be ensuring there are clear and protected paths for all data, and that it can move in predictable ways. “All of those algorithms and all the computing in the world are critically important,” suggests McCammon in conclusion, “but if the data gets mixed improperly, we’ll find ourselves dealing with more critical and catastrophic issues.” Through well-designed frameworks, Wind River hopes to help the automotive industry avoid any unwanted issues – especially those that can be described as critical and catastrophic.

This article appeared in the Q4 2016 issue of Automotive Megatrends Magazine.