On the Edge in Autonomous Vehicles: Part 2
Jumping the Hurdles
According to McKinsey, “AD and ACES are mutually reinforcing developments in the automotive industry, disrupting the automotive value chain [and] impacting all stakeholders.”6 Consequently, the coming changes will drive “seismic consequences for incumbent automakers and suppliers.”7 Changes that drive seismic consequences often face numerous hurdles, even as various forces work to clear them. Consider the following:
A key obstacle is the lack of a clear, reliable, and consistent regulatory platform. For example, McKinsey notes that though the U.N. Economic Commission for Europe joins several governments in actively drafting legislation for autonomous driving, exact requirements are not yet clear.8 Few if any legislative standards have been enacted, leaving the various industry players with a lack of scale, never certain whether a development that is accepted by one authority will pass muster with others.
Unquestionably, regulators face many serious issues. To what extent should autonomous vehicles (AVs) be granted exemptions to federal safety standards? Which specific standards can be loosened to allow thorough testing? With issues of data security and privacy, legislatures must consider the degrees to which data generated by AVs is accessible to vehicle owners, insurers, manufacturers, and other parties, and who has the right to resell this data.9 There are also issues of jurisdiction. For example, in the U.S., the federal government sets motor vehicle safety standards, but states are mainly in charge of registration, licensing, infrastructure, safety inspections, traffic law enforcement, and insurance and liability.10
Next consider the technical issues. Many different players, complex technologies, and capabilities are all moving at breakneck pace. The question becomes whether lawmakers can even begin to keep up with key developments — which means that innovation is held back due to ongoing revisions to safety regulations. AV advocates fear rules could become obsolete before they even come into effect.11
Like technology, legislation can be used or misused. While vague or missing legislation represents an obstacle, legislation can also become a key enabler. For example, the U.S. Congressional Research Service says that “on February 6, 2020, NHTSA [National Highway Traffic Safety Administration] announced its approval of the first autonomous vehicle exemption — from three federal motor vehicle standards — to Nuro, a California-based company that plans to deliver packages with a robotic vehicle smaller than a typical car.”12
Finally, the U.S. Air Force has emerged as an unlikely champion in the legislative arena by testing and then issuing its safety endorsement of a fully autonomous helicopter:
Like technology, legislation can be used or misused. While vague or missing legislation represents an obstacle, legislation can also become a key enabler. For example, the U.S. Congressional Research Service says that “on February 6, 2020, NHTSA announced its approval of the first autonomous vehicle exemption — from three federal motor vehicle standards — to Nuro, a California-based company that plans to deliver packages with a robotic vehicle smaller than a typical car.”12
Finally, the U.S. Air Force has emerged as an unlikely champion in the legislative arena by testing and then issuing its safety endorsement of a fully autonomous helicopter: The equipment is now available for military use sans pilot. This approval is ultimately meant to set the stage for civilian certification of the technology — including approval of autonomous flights over urban areas.13
ADAS, AD, and eventually ACES technologies offer a potpourri of attack surfaces. Many of these technologies are internet connected, and for purposes of coordination they also connect with one another. With so many access points, smart cars become highly vulnerable to hackers. At the annual Pwn2Own cybersecurity challenge, a team of two white hats was able to break into the infotainment system of a Tesla Model 3. From there, they ran minimal code of their own and soon took over operation. In other words, the car became theirs. Of course, the security flaw was reported to Tesla, and the car company soon issued a patch to fix that vulnerability.14
However, the point is that all systems attached to the internet are, at all times, targets for hackers. Regulators, academia, and industry leaders are taking a proactive approach to this problem. In the U.K., the Resilient Connected and Autonomous Vehicles (ResiCAV) consortium conducted a three-month study into the means for “detecting, understanding, and responding in real time to emerging cybersecurity threats across the mobility ecosystem,” releasing their findings in May of 2020.15
Another key hurdle is the uncertainty of communications standards. As Frost & Sullivan reports, “regulatory indecision is hampering automakers in North America and Europe. The U.S. government’s call for OEMs to comment on what existing/future technologies could be used for vehicle-to-everything (V2X) communications — dedicated short-range communications (DSRC), LTE cellular-V2X (C-V2X), or 5G New Radio (NR), among them — evoked widely differing responses. A similar exercise in the EU resulted in a majority of member states rejecting the Wi-Fi–based DSRC system.”16
Lacking clear regulatory direction, the global market is today split into two principal camps: proponents of DSRC and backers of C-V2X. Ultimately, Frost & Sullivan believes C-V2X will emerge as the most prominent industry standard. However, the group recognizes that most automakers will not commit until they determine their own preferences and strategies and see how regulations unfold.17
Of course, one of the greatest challenges in V2X communications in an ACES world is the achievement of low latency. As Srini Kalapala, VP of technology and network cloud for Verizon, explains, “It’s got to be low latency because you’re dealing with machines [moving] at high speeds. Things are being decided at a much faster pace. So we’re going to [move] from human-centric connectivity, delivering reliable connections for humans to do things, to now [needing] connectivity [at the speed of] machines.”
Click to listen to the full interview with Verizon’s Srini Kalapala The infrastructure leader of the second-largest carrier in the world talks about cloud infrastructure on the intelligent edge in this Forbes “Futures in Focus” podcast.
Another obstacle is the fact that ADAS, AD, and ACES capabilities must be tested and proven in real-world conditions. But to do so on just any highway or byway places the public at risk. Fortunately, the solution is at hand. According to a fall 2019 survey from Forbes Insights, 37% of OEMs and their suppliers “are making use of smart roadways for their connected vehicle testing, [with] another 42% intending to do so within three years.”18
Jeff Rupp, former chief technical officer and chief safety officer at the American Center for Mobility, describes the testing track his ACM group maintained as consisting of "500 acres less than an hour west of Detroit offering just about any characteristics you might need, from highway and rural roadways to urban environments.” The facility “includes features such as a 700-foot highway tunnel, a hazardous weather simulator, and a triple-deck highway overpass.” The University of Michigan Transportation Research Institute (UMTRI) offers a similar facility. Jim Sayer, UMTRI’s director and chief investigator, is tackling communications standards head-on. “Roadways will be smart, equipped with sensors so that traffic is directed based on real-time conditions (I2V), not historic trends,” he says. “It is important to demonstrate that cellular V2X and dedicated short-range communications (DSRC) can occupy adjacent channels in a real-world environment.”19
For both Rupp and Sayer, safety is paramount. “There’s an understanding among those working in this area that we cannot afford to make mistakes — as any missteps are likely to result in disproportionate pushback and delays,” says Rupp.
“What’s particularly important to appreciate about this future is just how many different disciplines and interests have to come into play for mobility to safely achieve its optimized future potential,” says Sayer. “So much learning and so many advances are taking place at such a rapid pace in interrelated fields, from materials sciences to AI and 5G. Even better understanding of the human factors within behavioral sciences can have a huge impact on safety [and effectiveness].” Creating partnerships with universities and research centers becomes essential. “Getting it right, bringing everything together, building in the needed safety requires a disciplined, systems-based approach aligning research, industry, and government,” says Sayer.20
Another key challenge for the industry will be finding the perfect balance between cloud-resident and edge-resident data and applications. According to Roman Pacewicz, chief product officer at AT&T, “The network itself is going to become a lot more intelligent, because basically it will no longer just move packets from point A to point B but will actually route application flows and enable this whole ecosystem to work. That has huge implications on network, network design, and also edge compute. Compute is going to have to come closer to where the processing and where the data applications are, and networks are going to have to become much more intelligent.”
Better-designed networks, along with low latency processing and broad connectivity, will bring the world many steps closer to what Verizon’s Kalapala calls “collective intelligence.” AI will be “extremely prevalent, [with] lots of decisions, lots of proactive conclusions being drawn from available data,” he says. As more data is drawn from more sensors on the vehicle, on the road, and all around, and then is processed within mere milliseconds, a state of collective intelligence can be achieved. It’s here, says Kalapala, that such a system can “drive all these deeper insights [leading to the] most profound outcomes — elevating your car to drive without having an accident.”
Click to listen to the full interview with Roman Pacewicz AT&T’s former chief product officer discusses the deep future of the cloud and its effect on society and the world in this Forbes “Futures in Focus” podcast.
A key challenge for ADAS, AD, and ACES is the fact that so much research and development is in flux. The amount of data being returned by sensors and analyzed leads to constant adjustments in the algorithms running in vehicle applications. This requires chips that are configurable while in use.
This also means that automotive companies are facing a major change in viewpoint. They need to start thinking like electronic systems companies, analyzing the system level, the chip level, and everything in between.21
Apparently, this mindset is already changing as chipmakers move deeper into working relationships with manufacturers. Chipmaker Hailo, for example, says it is working with leading OEMs and Tier 1 automotive companies to empower smarter edge and IoT devices. Key goals for advanced, specialized chips include not only advances in heat reduction, wattage, and latency but also in high-resolution segmentation and real-time object detection. The company says chips can now be designed and built to become highly compatible with neural networks, a trait that is a key enabler of autonomy. Accordingly, edge devices are becoming better than traditional solutions at running deep-learning applications at full scale with increased efficiency, effectiveness, and sustainability — all while significantly lowering costs, according to Hailo.22