'Level 2+': Making Automated Driving Profitable, Mainstream
Enhanced SAE Level 2 automated driving systems avoid the cost, complexity – and uncertainty – of Level 3.
Soon after SAE International released its J3016 “Levels of Automated Driving” standard in January 2014, engineers and product planners working in automated-vehicle development were probing the space between Level 2 and Level 3. New sensing and processor capabilities were poised to alter the lines of demarcation between L2’s focus on driver assist and the controversial (and tricky) driver alert and hand-back required by L3.
There was room to expand the ADAS (advanced driver-assistance systems) frontier. Those doing the probing believed SAE Level 2 could be effectively built upon, with greater functionality and safety potential, as new innovations became available. The terms “enhanced L2” and “Level 2-point-something” coined by engineers gave way to a more market-friendly “Level 2+.” And as 2021 begins, Level 2+ is the dominant ADAS trend, engineers and analysts tell SAE International.
“It’s clearly a big leap from Level 2 ADAS to [fully autonomous] Level 4, with leaps from L2 to L3 and L3 to L4 in between,” said Christophe Marnat, senior VP of the Electronics and ADAS division at ZF, which popularized the Level 2+ tag. “We started to talk about Level 2-Plus some years ago to show new systems that add safety and comfort features but always keep the driver in the loop. Because the more you remove the driver from control of the vehicle, in any conditions, the more systems quickly become very complicated. And expensive.”
Marnat sees a transition from Level 2 to Level 3 as “cost-prohibitive from an end-consumer perspective at this stage.” Developing an L3 solution requires substantially more computing power. It requires multiple (in some cases triple) redundancies in the vehicle’s sensor suite and related actuators, along with fusing huge amounts of data from multiple sensors. There are also limitations related to the HMI (human-machine interface) with L3, he argues, as the driver must come back into the control loop in a very short timeframe. And the regulatory framework for fully autonomous vehicles remains unsettled.
“It takes a very different system to accurately check if the driver is able to take back control. The vehicle must make the right decisions very quickly,” noted Marc Bolitho, head of engineering for ZF’s Electronics and ADAS division. “You start to get into the full hardware set you’d need for a Level 4 system. It’s one of the prohibitors right now for moving in that direction in the pass-car market.”
The industry’s interest in broadening the L2 space — at the risk of blurring some definitions — came as no surprise to the industry experts who mapped out SAE’s six-level AD taxonomy. They created J3016 (see chart) as a “living” document to be updated and modified to keep pace with technology and regulations.
Experts interviewed for this article said three factors forced the industry to alter course on its journey toward SAE Level 4 full self-driving systems, further encouraging the rise of Level 2+ “semi-automated” technologies and more pragmatic expectations. One factor was the COVID pandemic and the resulting “touchless economy” that curtailed, at least temporarily, the growing enthusiasm for robotaxis and ride sharing. Automated solutions to deliver food became the immediate focus.
The other two factors are the considerable economic constraint across the industry, combined with the strong regulatory push for Electrification, particularly in Europe and China. That dynamic, Marnat noted, “has changed priorities on what developments the OEMs are focusing on first. Electrification is a necessity and a key to survival.”
For ZF that means heightened activity on both Level 2 and Level 2+ systems. “This is where we see mainstream demand for a long, long timeframe ahead,” said Marnat, who keynoted SAE’s “ADAS to Automated Driving Digital Summit” event last month. “We support Level 3 and Level 4 as well with long range radar and lidars on the computing platform. But we believe the market is going to be focused on particularly Level 2-Plus for a long time,” he said. “There will be a few solutions for Level 3 and beyond, but we believe this will be a marginal portion of the market.”
The company’s engineers are developing new functionalities for the L2 space aimed at improving the existing driving experience and safety. These include adaptive merging for when vehicles are entering or exiting the highway, and various types of enhanced automatic emergency braking (AEB) aimed at improving vehicle-to-pedestrian safety and car-to-cyclist (and motorcyclist) safety. Also in the development program for production are interior-monitoring technologies to ensure driver attention.
The cost-vs-value ratio
Cost reduction is a paramount point made by OEMs and AV engineers. “Bringing in the enhanced ADAS functions, executed at high quality, at lower cost is a must,” according to ZF’s Bolitho. “This is where Tier 1s bringing complete turnkey solutions provide value — not only with sensing hardware and computing solutions but also the vehicle integration.” This is where he and Marnat see the industry moving, still within the “Level 2 bracket” that they assert is the mainstream.
In a 2020 interview with the author, Hyundai Motor North America’s chief safety officer Brian Latouf stressed the continued need for higher value in automated-driving systems, as new capability is added. “Hyundai continues to evolve our Level 2 ADAS systems with new vehicle enhancements to provide greater capability and improved radar camera sensor fusion technology,” Latouf stated. “We are also evaluating driver monitoring within our Level 2 systems to enhance the potential to allow a greater hands-free experience.” He said Hyundai continues to research the potential benefit that additional sensors, such as lidar, may provide in improving system capability, safety performance and driving experience.
GM global product development chief Doug Parks also highlights the importance of cost mitigation as the company further develops its (Level 2+) Super Cruise and next-gen Ultra Cruise systems for passenger vehicles. The goal, Parks asserted in a recent interview with The Detroit News, is continuing “to add self-assist capability without tens of thousands of dollars [in cost].
Other mobility-tech players are witnessing the trend toward expanding the SAE Level 2 space. "I'm seeing it with my OEM customers,” said Aptiv CTO Glen DeVos during a recent media backgrounder attended by AVE. “They're looking at everything they can do in a Level 2 system and enhance that system and basically bring it right up to the line where it is still a driver-assistance system, with a driver-in-the-loop, but really enrich the feature set that goes with that,” he noted.
“They can do that at a very nice price point and it's a very good margin option for them with the end customers. And it's a differentiator,” De Vos asserted. “We are definitely seeing that happen."
Modularity is key
Affordability for Level 2+ was a key driver in ZF’s development of a modular, scalable hardware-and-software suite now entering production with a major Asian OEM. The coAssist system uses cameras, radars and central ECU — no need for lidar, Marnat said — and leverages partner Mobileye’s EyeQ technology, to deliver functions including adaptive cruise, lane-change and lane-keeping assist, traffic jam support and traffic sign recognition.
Marnat describes it as “a plug-and-play solution that brings in the traditional suite of Level 2 and 2+ functions, priced at far below $1,000 [US]” and meeting projected Euro NCAP 2024 test protocols. He claims the quality of coAssist’s functional execution is high. “If you talk about adding the Level 3 functionality it’s a much higher price point than that,” he noted. “OEMs are aware of what the customer is willing to pay. The market for Level 3 through 2025 may not be as big as one might think.”
ZF has developed a modular range of L2+ systems including coAssist, coDrive and coPilot that allow vehicle manufacturers to upscale their functional offerings. The ability to add the company’s new Gen21 medium-range radar, 360° surround-camera perception and advanced processing can enable hands-free/feet-free, automated lane change and overtaking, automated parking, etc.) to meet OEM requirements across vehicle portfolios and price points within L2+.
Focusing on Level 2+ has enabled ZF to amortize fixed costs while enabling its engineers to learn new digital methods for testing and validation. “As we consider the cost sensitivity of the market and granularity of the [ADAS and AV] technology six years down the road, we believe the mainstream market is going to remain Level 2 and Level 2+ for a long, long time,” Marnat said.
Another angle: Honda leaps into Level 3 in Japan
While the industry is investing in Level 2+ ADAS production applications on a wide scale, OEMs continue to develop and test L3 vehicles. In November 2020, Honda Motor Co. received a type designation for SAE Level 3 automated driving from Japan’s Ministry of Land, Infrastructure, Transport and Tourism (MLIT). The approval enables vehicles equipped with an L3 version of the company’s “Traffic Jam Pilot” system to drive itself under certain conditions, such as in congested freeway traffic. Retail sales of an L3-equipped Honda Legend are expected by late March.
To promote commercialization of L3 automated vehicles, the MLIT partially amended Japan’s Road Vehicle Act last April. By the amendment, equipment for L3 driving was added to the list of motor vehicle equipment subject to Japan’s safety standards.
“The levels of automated driving (1-5) as defined by the Japanese government are based on SAE definitions,” explained John Dirrig, chief engineer, Corporate & Technical Communication, at Honda R&D Americas. The requirements in the amended Road Vehicle Act “set standards for the performance of the automated driving equipment,” he said. They include a requirement that the vehicle record and save data, and give law enforcement officers the right to inspect the recorder and request the submission of recorded data.
The MLIT also has “encouraged” (rather than mandated) a sticker be applied to the rear of the vehicle, indicating it is an Automated Vehicle. Of the on-board data recorder, R&D spokesman Dirrig explains that the L3 vehicle “must be able to record data to confirm operating conditions of the automated driving equipment such as the time when the equipment was turned on/off, the time when the equipment issued a handover warning, and the time when the driver fell into a situation where a handover was not possible.”
The recorder must be able to record this data for a 6-month period or 2,500 times, he added.