SAE clarifies autonomous driving level definitions

The Robot Report
SAE J3016 Levels of driving automation chart

The SAE J3016 Levels of Driving Automation is the official definition of autonomy level. | Image credit: SAE International.

The Society of Automotive Engineers (SAE) International announced some changes to the details of their six-level classification of autonomous driving capability. SAE International updated the way that they officially define the different levels while including new remote technologies and clarifications.

Enhanced definitions further refine the SAE Levels of driving automation

The definition of autonomous driving capability is a spectrum of functionality that starts with basic driver support functionality, all the way to the top level, which is complete driving autonomy. In clarifying the different levels, one of the new additions to the charts is that the SAE and ISO define the first three levels (L0, L1, and L2) as “Driver Support Systems,” while L3, L4 and L5 are used for actual “Automated Driving Systems.” Needless to say, no autonomous driving company has yet achieved SAE Level 5 autonomy, and many industry observers believe that Level 5 autonomy may still be decades away from becoming reality. However, the industry does believe that remote driving assistance will become an important step along the path to Level 5 autonomy, and thus, the new revision of the SAE chart reflects that inevitability.

The SAE J3016 chart is the official six-level definition of autonomous driving capability and was released in 2014 through the organizing bodies of SAE International and the International Organization for Standardization (ISO). The chart (see below), has been improved to include a clearer definition of driver support features, including lane centering, brake/acceleration support and adaptive cruise controls. It’s worth noting that a human is always driving for SAE levels 0-2. Automated driving isn’t a function until you get to SAE Level 3. In fact, all consumer automobiles currently on the road, including Tesla, are only at driving automation level 2. Tesla has had to backtrack recently on Tesla’s Elon Musk’s claim that Tesla would be at Level 5 autonomy by the end of 2021.

How to read the SAE J3016 chart when shopping for your next automobile

Most automotive manufacturers have developed and commercialized some form of SAE Level 0-2 autonomy as options in their 2021 model year vehicles. The technology is marketed now as options such as “Intelligent Cruise Control”, “Smart Cruise Control”, “Intelligent Lane Control”, “Lane-Keeping Assist System” or simply: “Adaptive Cruise Control”. All of these features help drivers to maintain safe control of their vehicles or more quickly react to a pending accident. Does it safe lives? Yes, and that’s a good thing. In addition, many automotive manufacturers are now delivering this functionality as standard equipment, rather than an add-on.

For example, Subaru offers its form of Level 2 autonomy as standard equipment on the 2021 model line up as the following standard features on the vehicle window sticker:

Waymo leads the way in Level 4 autonomy

Waymo is one autonomy provider who is currently operating a driverless taxi service at SAE Level 4 autonomy. However, you can’t purchase a Waymo Driver enabled vehicle just yet. Waymo One is their fully public, fully autonomous ride hailing service. The Waymo One service and the Waymo Driver enabled vehicles are still in development, but now anyone can take fully autonomous rides anytime they’re in Metro Phoenix. The company claims over 20 million miles of experience on public roads. However the systems are very closely monitored and they operate in a part of the world (Phoenix) where the sun is almost always shining and it never snows. This makes for ideal operating conditions and helps to reduce the risks for Waymo passengers.

Getting to Level 5 autonomy will require that the vehicle be able to drive at all times of the day, in all types of weather (rain, snow, sleet) and on all types of roads. It’s the last two conditions (all-weather and all road types) that present the biggest challenge to autonomous systems. Weather conditions like rain, sleet and snow reduce the effectiveness of on-board sensors. In fact, some weather conditions may completely blind a sensor, making it useless. If falling snow gathers of the front of a LiDAR unit, then it can no longer return 3D data for the vehicle’s perception engine. Likewise, driving on roads without markers (as on many rural roads) presents different challenges to the artificial intelligence onboard. Since the launch of the first Tesla vehicle, Tesla drivers have inadvertently been teaching the Tesla machine learning (ML) model how to drive. Using sensor data feedback and human driver inputs, Tesla uses it’s human owners as input to the ML models using everyday, on-road data. From that data stream, the Tesla ML-model learns from the choices that human drivers made in any given driving situation.

It’s noteworthy to compare the types of sensors on a Waymo Driver equipped vehicle to a Tesla vehicle. A Waymo Driver equipped vehicle has 29 cameras and other sensors around the vehicle, all providing an overlapping view of the world around the vehicle. Using sensor fusion, Waymo Driver can identify and avoid obstacles while computing the correct path of the vehicle. The types of sensors include vision cameras, LiDAR and radar. Compare that to Tesla with eight surround cameras and 12 ultrasonic sensors, in addition to forward-facing radar. Tesla does not employ LiDAR, and in fact Elon Musk believes that “LiDAR is a fools errand”. Figure 2 illustrates the different approaches to sensor allocation between a Waymo Driver configured vehicle and a standard Tesla model S. The additional sensing capabilities of the Waymo Driver configured vehicle is just one of the things that enables it to reach SAE Level 4 autonomy. [Editors note: figure 2 doesn’t attempt to identify every sensor location on the vehicle, however a Waymo Driver equipped vehicle carries an obviously greater array of sensor appendages than a Tesla vehicle]

Waymo Driver vs Tesla sensor map to vehicles

Figure 2. Waymo Driver includes additional sensors not included on a standard Tesla model. | Image credit: The Robot Report

Takeaways

The revised SAE levels of driving automation help to refine the milestones on the road starting from no automation through complete autonomy in passenger vehicles. With every automotive manufacturer and autonomous vehicle technology provider jockeying for a leadership role in the market, consumers have been left confused about what it means for a vehicle to be completely autonomous.

“The latest J3016 graphic is a living document. It will continue to evolve gradually as the industry and the technical standard J3016 itself evolves,” noted Jack Pokrzywa, SAE’s Ground Vehicle Standards Director. This is good news for the industry as a whole, and it helps to reign in the “one-upsmanship” that is currently taking place. Ideally, it also serves to educate the public about the limits of automation currently available in the consumer vehicle market while saving lives.

The post SAE clarifies autonomous driving level definitions appeared first on The Robot Report.


Source: therobotreport

Leave a Reply