By Angel Preston, Policy Director of Autonomous Vehicles and Internet of Things at Intel
Angel Preston, IntelAutonomous vehicle (AV) technology has the potential to improve roadway safety, save lives and transform personal mobility, offering broader access to mobility for the elderly and those with disabilities. It can also provide environmental benefits including reduced emissions and less traffic congestion. Innovation in AV technology is accelerating rapidly. During the recent Consumer Electronics Show, Amnon Shashua, president and CEO of Mobileye, an Intel company, revealed technological developments that show how close we’re getting to the promise of AVs. Policymakers in the United States now need to move quickly to develop regulations if they want to keep pace with the technical progress of the industry. Newly confirmed Secretary of Transportation Pete Buttigieg and staff have an opportunity to move things forward.
Mobileye develops both advanced driver assistance (ADAS) technology for human-driven vehicles as well as systems for fully autonomous vehicles. Mobileye’s approach integrates Road Experience Management™ (REM™) AV mapping technology, the Responsibility-Sensitive Safety (RSS) model , and two independent sensing subsystems based on world-leading camera, radar and lidar technologies. This trinity of innovations enables an AV that is orders of magnitude more capable than human drivers.
REM Mapping
Mobileye found that many high-definition digital maps were overly specific while lacking features AV systems need to understand and contextualize their environments. “The current state of AI can detect road users at fidelity approaching human perception,” Shashua says. But, he says, it’s not realistic to think the in-vehicle AI can do that and, in real-time, “understand the many complexities of the driving environment ¬— lanes, crosswalks, signals, signs, curves, right-of-way, turns.”
Starting in 2018, Mobileye began crowdsourcing – using information sent from vehicles already equipped with Mobileye ADAS technology – to automatically map the environment in ways useful to AV systems. Cars including those mad e by BMW, Nissan and Volkswagen have mapped nearly one billion kilometers of roads to date. By 2024, Mobileye expects that level of information to be transmitted daily utilizing only a trickle of data (10KB/km), easily supported by existing cellular networks. This revolutionary AV mapping technique produces critical information such as drivable paths, lane association and priority, association between traffic lights and crosswalks, stopping and yield points, and common speed. This information is crucial to realizing AVs’ safety promise and is not typically found in traditional HD maps.
The AV maps are also providing Mobileye with an enormous amount of relevant road data – so much that they will be able to start testing AVs in four new countries without needing to send specialized engineers or mapping vehicles to those locations. The company will instead send pre-equipped vehicles to local teams that support Mobileye customers. In a matter of days, those vehicles will be driving autonomously with a safety driver. This approach was used in 2020 to enable Mobileye AVs to start driving in Munich and Detroit within a few days of touching ground.
RSS Model
In 2017, Intel and Mobileye proposed a technology-neutral, open and transparent approach for safety based upon the Responsibility-Sensitive Safety (RSS) model, which helped shape the forthcoming IEEE P2846 standard. RSS and the IEEE standard recognize that there is no such thing as perfect safety. Rather, the safety of AVs is based on the ability to make reasonable worst-case assumptions on what to expect of other road users, just as humans do. By doing so, AV makers can deliver a greatly increased level of safety within the bounds of what is reasonably foreseeable, consistent with a risk-minimizing approach that NHTSA has taken in a variety of other areas .
Radar/Lidar
Mobileye’s two independent sensing systems, one based on cameras and the other on radar and lidar, are not just complementary; they are redundant. Each can enable autonomous driving without the use of the other. Cameras with Mobileye’s industry leading computer vision algorithms remain the primary sensor in the AV. Simultaneously, radar and lidar provide True Redundancy™ to the camera sensing and perception system.
However, if the benefits of AVs are to be truly realized, we must deliver not only independent and redundant sensors, but also a self-driving system that is economically scalable and able to drive anywhere a person would be expected to drive.
Mobileye and Intel are developing two novel sensors, a silicon photonics-based lidar and a software-defined radar. These promise to deliver significant performance enhancements over existing radar and lidar systems at greatly reduced cost. This will democratize the AV revolution, making safety and self-driving more broadly available throughout our society.
Policy
Before these technological breakthroughs can benefit society, we need a U.S. policy breakthrough. The National Highway Traffic Safety Administration (NHTSA) should develop Federal Motor Vehicle Safety Standards for AVs so the industry has clarity on the expected performance capabilities of AVs and can ensure AVs meet those requirements before consumers start adopting them broadly . We believe a risk-minimizing approach such as RSS or IEEE P2846 can facilitate this policy breakthrough.
Revolutionary technology is making the future of autonomous driving safer. Now is the time to define performance requirements for AVs so that the nation can realize the numerous benefits of autonomous vehicles.