Skip to content

Autonomous drive tech advances – can regulations keep up?

Freddie Holmes speaks to Elektrobit’s Walter Sullivan about how autonomous driving technology has ramped up since the supplier set up shop in Silicon Valley

In January 2015, Elektrobit (EB) established a Silicon Valley-based Innovation Lab as a means to ramp up developments in connected and autonomous vehicles. At the time, Walter Sullivan, Head of Innovation at the San Jose facility, said the plan was to “collaborate and engage” with the ecosystem of companies in this space.

Two years down the line, the small EB team has made serious progress. “This summer we built a self-driving car at our Innovation Lab,” Sullivan tells Megatrends. “We started from an existing platform, but we replaced some of the sensors and we now have an automated driving development platform.”

In addition, EB has been carrying out research on sensor technologies, and has been working on a joint project with a US-based OEM to collect operational data for that company’s automated driving systems.

Nurturing concepts

Today, the Innovation Lab is where initial concepts and ideas are formulated, but most of what Sullivan calls the “heavy lifting” is then coordinated from EB’s Detroit and Finland headquarters. “Not every idea starts life in the Silicon Valley office, but the point of having an innovation centre in Silicon Valley is twofold. One is that pretty much every vehicle manufacturer has a research office there, so part of my responsibility is to maintain communication with those guys, understand what they’re doing, and see if there are ways Elektrobit can work with them,” explains Sullivan.

The second key benefit is that EB can essentially scout out new technologies coming from the innovative and fast-paced Bay Area, or to see which start-ups are attracting venture capitalist money. “We’ve been in contact with a number of new LiDAR manufacturers, and with a couple of new high precision positioning companies in the Bay Area that are augmenting GPS,” notes Sullivan. “It’s obvious, but an autonomous vehicle has to know very precisely where it is, to within centimetres. GPS in a traditional car today is only accurate to roughly five metres, and that can be a problem for driverless cars.”

The time it takes for a concept to enter a full production programme varies, and there is no defined process to dictate whether it is two weeks, two months or two years. For example, EB’s project with the OEM was prototyped in EB’s Silicon Valley office for just under a month before it was handed off to a separate team in other offices. “The amount of work for this project was more than my little team can do, so we ramped up a team in a couple of other offices and handed it over,” explains Sullivan. “We’re just a very small team in Silicon Valley, and I’m looking to expand the team by adding people with experience in autonomous driving or robotics, and in basic vehicle networking.”

An autonomous vehicle has to know very precisely where it is, to within centimetres. GPS in a traditional car today is only accurate to roughly five metres, and that can be a problem for driverless cars

Generally speaking, says Sullivan, setting up with bricks and mortar in Silicon Valley has proven a welcome addition to the company’s wider operations in autonomous and connected driving software.

Faster, better, stronger

Having been met with a high degree of scepticism, the idea of autonomous driving is now deemed an inevitability by most. Since the Innovation Lab’s establishment at the start of 2015, there has been a significant rise in activity from all stakeholders, from established suppliers and OEMs to up-and-coming start-ups. Regulatory bodies have begun to allow for pilot tests on public roads in several cities and states, in the US and around the world, in an effort to provide driverless systems with real-world experience.

Various OEMs have announced plans to launch vehicles with varying degrees of automation. On the SAE scale, level four automation describes a vehicle that can drive itself in practically all situations, whereas level five vehicles do not even require a human behind the wheel.

In October 2016, Tesla announced that all its vehicles would have the required hardware (sensors) for ‘full self-driving capability’ that would be ‘substantially’ safer than a human driver. At CES in January 2017, Audi announced that it plans to launch a level four autonomous vehicle in 2020. Mercedes-Benz also announced at the event that it would launch a car featuring self-driving artificial intelligence ‘within 12 months’. BMW plans to launch a ‘fully autonomous’ road car in 2021, with Ford also set to roll out a fleet of autonomous ride share vehicles that year.

“We may or may not see some limited availability of Level 5 autonomous driving in the first half of the 2020s, but we’ll definitely be at Level 4,” muses Sullivan. Such optimism, he believes, has been triggered by some key recent technology shifts, notably increased computing power and efficiency.

Creating a vehicle that can drive safer than a human requires considerable computing power. Most autonomous research vehicles on the road today still feature a rack of laptops in the luggage area, but computing power has improved dramatically in recent years, and has facilitated the integration of greater driverless capabilities.

Sullivan highlights the work of companies such as Nvidia and Intel. “They’ve been reducing the size of compute modules, and Nvidia in particular has been cranking up the compute cycles to levels where we can have a reasonable set of functionalities,” he explains. “That’s been a big change.” Nvidia’s latest system on chip (SoC) computer, dubbed ‘Xavier’, can perform 30,000 TOPS (trillion operations per second) and will be used as the self-driving computer for several OEMs. It can track eye gaze, and even lip-read driver speech with around 95% accuracy.

In addition to computers becoming more capable, Sullivan adds that they are also becoming more efficient in terms of electrical consumption. This reduces the amount of heat produced by the computer, which ultimately needs to be expelled from the car. “Heat generation is a problem, particularly when it comes to getting the heat out of the car so that passengers can be comfortable,” he explains. “The car could handle the heat, but those in the vehicle wouldn’t like it.”

The other change that has fast-tracked the development of vehicle automation is the falling cost of camera, radar and LiDAR sensors, the latter of which is deemed vital for level four/five autonomous driving. The LiDAR sensor used in Google’s first prototype test vehicle reportedly cost around US$75,000 per unit, but Waymo – the recently spun-off self-driving division of Google – announced in January 2017 that it had achieved a 90% decrease in the cost of LiDAR sensors in the space of 12 months.

Today, we don’t know what rules we’re going to be operating under. It’s somewhat counterintuitive, but the faster regulation can move, the greater the degree of certainty

Other start-ups, such as Magna-backed Innoviz, plan to launch LiDAR sensors for US$100 per unit.

“We’re really starting to see cost reductions in LiDAR coming,” says Sullivan. “We need a device that’s in the US$200 range because there could be four of them on the car – you can’t have an US$8,000 sensor if that’s the case.”

Regulation – a ‘wild card’

Contrary to popular belief among many consumers, development of the driverless car is not being held back by technology. Today, the primary challenge is the speed of development of road safety regulations, and securing the necessary support from national and regional governments.

In the US, brands need to prove that their autonomous driving vehicles are safe to test on public roads by submitting supporting data to the government. Numerous licenses have been granted across various states so far.

In September 2016, the National Highway Traffic Safety Administration (NHTSA) published a 112-page policy on automated vehicles, a document that was issued as ‘agency guidance’ as opposed to a firm rulemaking. The idea was to speed up the creation of an initial regulatory framework by issuing guidance on how to deploy autonomous vehicles and model state-by-state policies. Shortly after, in December, Michigan’s Governor signed the SAVE Act, which allows for free operation of autonomous vehicles on any road in the state. In January 2017, NHTSA then published an advisory best practices document with the aim of further improving how autonomous vehicles are tested, and ensuring they are deployed safely.

Sullivan is troubled by the speed of regulation to support the roll-out of autonomous vehicles, which he feels is moving too slowly. The problem, he points out, is that until firm regulation is in place, the industry is unaware of the requirements to which it will be working.

“Today, we don’t know what rules we’re going to be operating under. It’s somewhat counterintuitive, but the faster regulation can move, the greater the degree of certainty,” he concludes. “There’s the corollary that if you can get there first, the regulation will effectively bend around what exists. That might also be true, but right now I would say regulation is still a wild card.”

 

 

This article appeared in the Q1 2017 issue of Automotive Megatrends Magazine. Follow this link to download the full issue. 

Welcome back , to continue browsing the site, please click here