As the auto industry and its suppliers, along with technology developers and startups, recalibrate for what appears to be a new and more pragmatic phase of automated-driving expectations, industry conversation is turning to contextual applications and managing consumer perceptions. Concentration on SAE automated-driving Level 4 in particular is accelerating as developers embrace operational design domain (ODD) as an opportunity rather than a limitation — and SAE Level 3’s driver “handoff” complications are muddying implementation and understanding of Level 3 advanced driver-assistance systems (ADAS).
It was against this backdrop that Austria-based automated driving initiative The Autonomous conducted its Main Event conference in Vienna in late September with the stated theme of “Making safe autonomous mobility a reality.” Representatives of some 200 companies convened to discuss Level 4 aspects such as genuine first-use cases, the role of artificial intelligence and how to guarantee Level 4 system safety. Just prior to the event, speakers from the “Challenges for Level 4 Safe Automated Driving” panel participated in a virtual roundtable with international reporters, including SAE Media.
Safety uber alles
The Autonomous initiative cited a McKinsey forecast valuing the automated-vehicle market at $1.6 trillion by 2030 and also expressed faith that Level 4 automation can reduce fuel consumption by 15-20%, cut emissions and provide new mobility opportunities for the elderly and those with disabilities. But to achieve those payoffs, the Level 4 Safe Automated Driving panel stressed that reliable and unassailable safety for automated driving is crucial to winning public confidence.
Peter Schaefer, EVP and CMO of the Automotive Division at Infineon, stressed that all manner of “ingredients” must be assembled to “make a system dependable.” He said that will come from a combination of technology, consumer trust and dependable electronics. “The other underlying topic to solve is to make the systems safe, secure, and highly available,” Schaefer added at the conference. “If we have a wonderful Level 4 system and it's not available, consumers will ultimately lose trust in new systems.”
Stefan Poledna, CTO of TTTech Auto, added that the ultimate path to public trust is to demonstrate that automated vehicles are safer than human drivers. Current system abilities are good, the speakers agreed, but to get to 100% reliability “might be a long journey,” Schaefer conceded. Nonetheless, Indu Vijayan, director of Product Management at AEye, a developer of software-defined lidar sensing, said she believes Level 4 automation will be available for private vehicles in a decade or sooner.
Hardware, software and systems
Asked whether hardware or software improvement is more crucial to advancing high-level vehicle automation, the panelists weren’t in agreement, effectively saying that both are important. “It’s a very holistic approach to get a system that is dependable,” stated Infineon’s Schaefer.
Vijayan offered data about how perception-system performance is advancing to provide more sophistication — and margin for safety. She said that 120 to 160 meters (394 to 525 ft.) is now the target for “seeing a small object at long range” for highway-oriented Level 4 automated driving systems. Several automakers currently market “hands-off” driver-assist systems designed largely for highway driving, but all are Level 3 category — meaning that although the system will accelerate, decelerate and steer the vehicle, the driver must at all times be prepared to take control if necessary.
This distinction is one that Vijayan said is important, noting that system limitations need to be clearly defined for end-users. A recent study released by the Insurance Institute for Highway Safety (IIHS) supported this point, finding in a survey of 600 users of various automakers’ ADAS systems that those users had “poor understanding of the technology’s limits,” according to IIHS President David Harkey.
The panel agreed that safety considerations should not be dictated by any specific strategic choices about hardware, such as chipsets, or software — and that a guarantee of system safety shouldn’t be regarded in a competitive sense but in a collaborative framework. “It’s very clear that we need to have redundancy; no single chip, sensor or software component can do it alone,” said TTTech’s Poledna at the conference. “You need to make sure that any single failure can be mitigated.”
Commercial implementation pulling ahead
At The Autonomous’ Main Event conference, there also was discussion regarding the increasingly accepted premise that commercial-vehicle business cases for Level 4 automation are pulling ahead of potential personal-vehicle implementation. “I want to say yes,” asserted Magnus LilJeqvist of Volvo Trucks. “Fully L4 is the ‘golden nugget,’” for commercial vehicles, he said, adding that there are profitable business cases for Level 4 automation “even in limited ODDs (operational design domains).”
Hakan Schildt, senior VP at Traton Group, sees Level 4 for trucking progressing in a “phased progress” from limited-ODD service to A-to-B routes. “We can scale with profitability,” he said at the conference. “That scaling leads to rapid ODD expansion. It’s very much a scaling business.”
Annie Lien, deputy CEO at artificial-intelligence developer VinAi, admitted to a different perspective than many truck-industry proponents of Level 4. She said that Level 4 operation for personal vehicles will take “probably a lot longer [to implement] than people realize. She forecasts significant L4 adoption will come “a lot later” than 2030. “We’re going to be in Level 2 for quite a bit,” she continued. “L3 is problematic, I think we’ll all agree. We’re going to be drivers in the car for a long time,” she said, adding that driver-monitoring systems, one of her company’s products, are critical and also will be necessary for some time to come.Continue reading »