Don’t worry, cracks in the autonomous road are just disrupters!

Thilo Koslowski, vice-president, analyst and founder of Gartner's Automotive Practice, has said that, ultimately, the car will become a node in a much bigger network. This will create new intermodal opportunities and become part of a smart city environment.

Car-to-car infrastructure is a huge new opportunity being enabled by connected vehicles – and will see the most investment and strategic decision-making.

In 2016, connected-vehicle features will become a critical buying consideration for the average consumer in mature markets.

Koslowski has placed autonomous vehicles at the peak of the hype cycle. Right now, he says, consumers have inflated expectations. Now, we have to reset those expectations. You won't be able to sleep while your car drives you to work for many years. He also promised that self-driving cars won't linger too long in the trough of disillusionment.

He cited a recent Gartner consumer survey. When asked: ‘Do you want a car that can drive itself?’, 27% were very interested, and 41% said, "maybe".

He reiterated that fully autonomous vehicle will be on the road by 2020, with significant penetration by 2030. "By 2020, 10% of today's vehicle owners in mature markets will give up vehicle ownership for on-demand access," he says.

Autonomy is now a given. You can already see it in action on highways and city streets. So it's very easy to forget that really major problems have yet to be solved. Luckily, we had plenty of sessions to remind us of the challenges that remain.

One of the greatest challenges is the machine-to-human handoff. It's unreasonable to let drivers do other tasks and then expect them to take over quickly in an emergency.

"When you tell people we'll drive for you but you have to supervise the car at all times it's unlikely to be perceived as real customer value," says Jim Mazurek, senior vice-president of automotive sales and business development at Neusoft.

Olaf Preissner, head of UX automotive and innovations at Luxoft, notes: "When systems get more reliable, takeover skills get worse." This is an established problem in commercial aviation, and pilots are required to take training to maintain their skills. Will drivers be willing to do the same?

"The system has to deliver right amount of engagement at right time to keep driver's mind from wandering," Preissner says.

It's not only a human-machine interface (HMI) problem. It's also a people problem – and everyone is different.

"We can design the tech but how do you tune it to deal with the behaviour of individuals?" asked Mazurek.

Now that automakers have firmly planted in everyone's mind – including National Highway Traffic Safety Administration’s (NHTSA's) – that autonomous driving will progress through several levels before true driverless driving, experts have started to talk about skipping level 3.

"Level 3 is the tricky one. You have a thing that encourages you to delegate driving but you can't, because if something happens, it's your fault. Level 2 systems are amazing and will radically improve the commutes of most Americans. We would like to leapfrog from 2 to 4," says Anders Tylman-Mikiewicz, general manager of the Volvo Monitoring & Concept Centre.

Meanwhile, Chris Heiser, CEO of Renovo Motors, says the industry should stop dithering. Autonomy may not be perfect; self-driving cars may still get involved in crashes. Yes, there's a level of risk, he says, but they'll still be better than human drivers.

Roger Lanctot, associate director of the global automotive practice of Strategy Analytics, posited that fully autonomous vehicles, such as the Google cars without steering wheels or pedals, might be best deployed in places like Las Vegas, where automotive and pedestrian congestion are expected to soon exceed capacity.

In the race to autonomy, someone forgot about security. Okay, not someone – almost everyone.

Back in September 2014, security researcher Chis Valasek told TU-Automotive that modern cars were loaded with "attack surfaces." But it wasn't until July 2015, when Valasek and fellow researcher Charlie Miller demonstrated they could disable the accelerator pedal on a Jeep – doing 70mph on a highway – that the industry started to take security seriously.

It's serious now. Most experts now agree the need for better security. They also acknowledge that securing the connected car might take revamping every aspect of the design and manufacturing process.

It's no longer necessary to try to drum up awareness, says Joe Fabbre, director of platform solutions, Green Hills Software. Even though over-the-air (OTA) can be used to issue quick security patches, as we saw with Tesla, the industry should not be lulled into a false sense of security, he says.

Most security defects are the result of software bugs, according to Fabbre, and some bugs are security vulnerabilities. If an automobile meets the software-industry standard of one to 25 defects per 1,000 lines of code, over five years, its software could have 2,500 discovered defects – and 7,500 that are not discovered but could have the potential for zero-day attacks.

He proposes a solution that separates safety-critical components in the system from untrusted code while enforcing strict access control.

This isn't always possible, says Chip Goetzinger, senior manager for vehicle connected services at Nissan. For example, because the infotainment system is often the best way to deliver information about the car's systems to the driver and to deliver OTA software updates, the distinction between critical systems and infotainment is starting to blur.

Standardisation of platforms could relieve some of the security pressure. Matt Jones, head of future infotainment and senior technical specialist at JLR, highlights the GENIVI project, which is about to release its next platform. "It makes sense for organisations to standardise the bits customers don't pay for," he says.

In addition to two-thirds of an IVI stack, GENIVI is working on Automotive Message Broker, a framework for accessing vehicle information via an application instead of directly. The plan is to extend it to control the car's heating, ventilation and air conditioning (HVAC) systems.

Many experts suggest the auto industry needs to stop dithering about standards. In the first place, there are security standards already that could be applicable to automotive. The problem with standards consortia in general, though, Goetzinger says, is that most companies are already too invested in their own systems. The common response to attempts at standardisation is, "Standards are great. Let's use ours!"

Andrew Poliak, global director of business development for QNX Software Systems, says automakers will need to rely on their vendors for security as standards are developed – a process that could take 20 years.

Still, security is always a race against hackers, says Dirk Reimer, vice-president of sales and marketing for Telit Automotive Solutions. What makes the difference, he says, is ‘different’ is how much effort companies put into assessment and validation to make the frequency of hacking as low as possible.

Jonathan Allen, a principal and director of Booz Allen Hamilton, is also executive director of the Auto ISAC, an industry group that will share information about threats and attacks. He gives kudos to automakers for creating the ISAC before a catastrophic attack.

He adds that disaster planning is crucial – and needs to be continual. After the Jeep demonstration hack, he says: "Every OEM should have started their incident response planning that afternoon."

Koslowski delivers a chilling prediction: by 2019 two automotive companies will be fined for vehicle software design negligence that results in inconsistent technical performance or cybersecurity attacks.

He says: "Automakers can no longer say it's not their problem. OEMs must create an end-to-end technology approach that takes security into account."


Leave a comment

Your email address will not be published. Required fields are marked *