This year our friends at the United States Department of Transportation designated October as Pedestrian Safety Month, so we took the opportunity to focus on how automated vehicles can help improve the safety of vulnerable road users (VRUs, a common collective term for pedestrians, cyclists and other non-car road users). And we weren’t alone: our members stepped up as well, sharing fascinating insights from their work to help explain how their technologies can help improve the safety of pedestrians and cyclists. In hopes of highlighting just how much effort is going into addressing this problem, we’re bringing as much of this work together here into a one-stop blog post covering the latest in VRU-focused automation.
First, however, it is important to understand the problem our members are working so hard to solve. PAVE members are making great strides: the Governor’s Highway Safety Association has been a leader in studying pedestrian and cyclist safety, in partnership with the Pedestrian and Bicycle Information Center.
GHSA’s most recent report on this topic, which looks at 2019 preliminary data, projects that 6,590 pedestrians were killed on American roads last year, a five percent increase from 2018 which, in turn, was up 3.4% from 2017. Once verified, that would make 2019 the deadliest year for pedestrians in this country since 1990, showing that these numbers are trending in the wrong direction. Even as a percentage of all road traffic fatalities, pedestrian deaths have been on the rise for some time, climbing from 12% in 2009 to 17% in 2018… a share not seen since 1982.
We had GHSA’s Pam Fischer on one of our weekly virtual discussions, and she helped us dig into this data to understand what is causing these deaths. She pointed to a wide variety of factors, from infrastructure and street lighting to the growing size of vehicles, noting that 76% of pedestrian fatalities in 2018 took place in the dark, 59% took place on non-freeway arterials and 47% took place in just five states (Arizona, California, Florida, Georgia and Texas). Booming sales of trucks and SUVs, rising driver distraction and persistent problems with intoxicated driving seem to be factors as well.
With this overview of the problem in mind, we turn to the opportunities to address these issues in the short term using driver assistance systems, which was the subject of our first October panel (which you can watch above). We were thrilled to be joined by Ken McLeod, Policy Director at the League of American Bicyclists, who gave us a cyclist’s-eye-view of the problem. Representing PAVE’s members were Rini Sherony, Senior Principal Engineer at Toyota’s Collaborative Safety Research Center, and Mike Walters, VP of Product Management for the components business at FLIR Systems.
Sherony explained that Toyota looks closely at real-world data when it develops new safety systems, and the troubling trends on VRU safety numbers contributed (along with stricter Euro NCAP standards) to Toyota adding pedestrian and cyclist detection to its latest Safety Sense suite. Her hope is that as these systems achieve higher penetration rates we will see the VRU data start to improve, although she cautions that some happen in complex situations that will require more advanced technology to prevent.
One of those more advanced challenges is preventing the large percentage of pedestrian fatalities that take place at night, and FLIR’s Mike Walters described how thermal imaging could address this challenge directly. Whether by augmenting the driver’s situational awareness, like a “spotlight” system already offered in some European luxury vehicles, or by automatically stopping using a thermal-powered AEB system that shows real promise in development testing, Walters argued that thermal imaging was key to the next step in pedestrian-aware ADAS.
Our second virtual discussion went beyond driver assistance to look at the fascinating ways in which fully autonomous vehicles are being prepared to achieve higher levels of VRU safety in the future. With guests from influential companies in the AV world — Mobileye’s Gaby Hayon, Cruise’s Nadia Anderson, and Sam Anthony, who founded a company called Perceptive Automata focused on solving problems related to AV pedestrian safety — this conversation highlighted a wide variety of approaches and a deep commitment to next-level safety.
Dr. Hayon’s comments touched on a wide variety of factors that go into Mobileye’s pedestrian safety efforts, some of which are summarized in a subsequent company blog post. From object detection to stereoscopic analysis, from pixel labeling to what the company calls “VIDAR,” Mobileye draws on its industry-leading experience in computer vision for automated driving (Hayon holds pedestrian safety-related patents dating back to 2004!) and marries that expertise to its Responsibility-Sensitive Safety framework.
Cruise’s Dr. Anderson described the importance of pedestrian safety to the company, which is rooted in its plans to deploy fully driverless vehicles in the challenging downtown of San Francisco, and explained the dizzying array of challenging scenarios that this presents. Highlighting the fascinating work her colleagues are doing to leverage motion capture technology developed for videogames, Anderson noted that a wide variety of techniques and technologies are required to handle the sheer variety of VRU-related challenges you see somewhere like San Francisco.
And Dr. Anthony described the work Perceptive Automata is doing, bringing insights gleaned from behavioral neuroscience into the world of machine learning to surface lessons that will help AVs keep pedestrians safe. Though that means Perceptive’s technology can interpret a pedestrian’s intent based on something as seemingly innocuous as how they hold a bag, he argues that Perceptive is not a “prediction” company in the traditional sense but rather a human understanding company. That means understanding everything from how people act at different times of day to what we can learn from the 2018 Uber autonomous test vehicle crash in Tempe, Arizona, and you can tell that Sam lives and breathes the challenge of working toward a future where pedestrian deaths are a thing of the past.
That brings us to our third panel, which looks beyond the core autonomous driving technology stack to understand the additional safety layers that AV developers can bring to their vehicles. Featuring GHSA’s Pam Fischer, Nuro’s Public Policy Manager Greg Rogers and !important Safety Technologies’ Bastien Beauchamp, this wide-ranging conversation touched on a broad variety of potential solutions to the problems Fischer describes in GHSA’s data.
Beauchamp’s company develops technology that provides one of these additional safety layers in the form of a smartphone-based beacon that can inform an AV that a pedestrian is present even if they can’t be seen by the perception stack. Calling this technology a “digital seat belt for the modern era,” he points out that many pedestrian deaths are caused by situations where drivers are not responsible and suggests that using connected technologies is one way to prevent deaths that AVs might not be able to prevent.
Rogers highlighted a variety of steps Nuro has taken to improve VRU safety, starting with the size of the vehicle itself: by developing a smaller, occupantless autonomous vehicle, Nuro hopes that the worst-case scenario would never be as bad for one of their R2 delivery bots as it would for a larger vehicle. Of course they aren’t stopping there, adding capabilities like thermal imaging to help detect pedestrians and even an exterior airbag to provide even more protection in a worst-case scenario. These steps, along with thorough development and validation of its core autonomous driving technology, allow Nuro to take the ultimate step of testing their R2s with no backup human safety driver in confidence.
Of course, there’s a lot more that our members are doing to leverage automated vehicle technology to improve VRU safety than we could highlight in our virtual discussions. Argo recently published a blog post on its approach to ensuring pedestrian and cyclist safety, highlighting both its technology and its engagement with cycling groups like the League of American Cyclists. Earlier this year, Waymo published a delightfully nerdy blog post showing how its automated data augmentation improves lidar point clouds of pedestrians and cyclists. Speaking of lidar, laser imaging pioneer Velodyne wrote its own blog post showing how its technology can improve VRU safety, particularly in the low-light conditions where so many pedestrian deaths happen.
After decades of improving safety for vehicle occupants, the crisis in VRU safety is clearly energizing both existing players in the automotive space, new startups and the next generation of AV developers to focus on the unique challenges related to this problem. Though our members caution that no one system or technology is a “silver bullet,” and that work must continue on a variety of fronts to make the necessary progress, we are incredibly proud of all the work they do to make the roads of tomorrow safer for everyone. We hope that this overview of the work they are doing, pushing the boundaries of technology forward, helps illustrate why. If you’d like to learn more, we encourage you to click through the links here for more details on this important work.