Despite the market challenges posed by COVID-19, the long-term outlook for automated vehicles (AVs)—many industry analysts believe—remains positive.VX News share's here with permission, an excerpt from the Reason Foundation's Surface Transportation Innovations newsletter in which author, Marc Scribner, analyzes the impact COVID-19 has had on developer plans for AVs including the implications for shared mobility, machine learning algorithms, and investor confidence. Scribner asserts that potential benefits to mobility and safety presented by AVs will outshine the current uncertainty.
Significant advances in vehicle automation have occurred in recent years as upstart Silicon Valley developers were joined by traditional automakers in an effort to produce what the Society of Automotive Engineers (SAE) terms Level 4 automated driving systems. These are what are popularly called driverless cars, enabling a vehicle to be directed by computer for the entire duration of a trip without any human driving responsibility—albeit subject to a limited types of roads and weather conditions. Then came Covid-19, leading a number of developers to push back their estimated deployment dates in the face of problems stemming from the pandemic. These delays are in part the result of the financial liquidity problems plaguing businesses generally, but are also caused by health-related business model risks, reduced ability to reliably train automated driving systems in real-world traffic conditions, and broad uncertainty surrounding post-Covid-19 travel demand. Understanding all of these factors can provide a better picture of when and how enthusiastic development may resume.
GM’s automated vehicle unit Cruise announced layoffs. Velodyne Lidar, which manufacturers a type of sensors used in most automated vehicles, did the same. Uber shuttered its artificial intelligence laboratory in the course of downsizing a quarter of its workforce. Lyft said its automated vehicle (AV) division was impacted by its decision to lay off or furlough more than 1,000 employees, without providing specific figures. Once-promising automated truck developer Starsky Robotics, which laid off nearly 90% of its staff in November 2019 amid growing financial difficulties, shut down and began liquidating its remaining assets in late March. This was accompanied by a melancholy message from CEO Stefan Seltz-Axmacher predicting the start of a five-year decline in AV investment and development.
Given the proprietary nature of these technologies and the intense secrecy that characterizes the industry, it is difficult to evaluate Seltz-Axmacher’s gloomy forecast. Intel’s MobilEye, in partnership with Volkswagen, has said its 2022 deployment of self-driving taxis in Israel remains on track. Ford and its partner Argo AI have pushed back the launch of their planned automated taxi service from next year to 2022. Others have yet to publicly adjust their estimated deployment dates. However, it appears that at least a one-year setback is likely for four reasons.
First, investors may tend to be more conservative until the pandemic passes and economy rebounds. This means risky, multi-billion-dollar bets on perfecting a Level 4 automated driving system over years with no revenue are less likely, especially for traditional automakers that can turn to more-proven business lines. Most companies are at least publicly optimistic about the limited impact of Covid-19 on their AV research and development activities, but investors will ultimately drive the outcome.
Second, many of these developers were building their initial AV business models around centralized fleets of self-driving taxis. As with all shared transportation, health concerns loom large. Will customers be willing to enter a vehicle that may have just transported an infected passenger? A deep cleaning following each use to ensure no viral transmission undercuts the economics of this business model. A May 2018 study in Transport Policy developed a low-cost Level 4 vehicle automation scenario where operating costs on a passenger-mile basis of a single-occupant self-driving taxi fall below those of a self-driving transit bus. However, the team of Swiss researchers highlighted cleaning as one of the uncertain costs that could make or break this prediction. Unless affordable cleaning technology is developed to automatically and quickly sanitize vehicle cabins after each use, developers may at least temporarily pivot to last-mile cargo delivery, such as the Level 4 vehicles from Nuro that are currently delivering groceries in Houston.
Third, with travel down on streets and sidewalks, companies are facing new challenges with machine learning. Contrary to popular narrative, these algorithms are not the brainy artificial intelligence from science fiction; rather, they are pattern recognition technologies. Abnormal traffic patterns are not very useful in teaching an automated driving system to safely drive in normal conditions. While minimizing public health risks was the primary reason for developers such as Waymo, Argo AI, and Cruise deciding to temporarily suspend public road testing, the Covid-19 travel decline likely played a role. They continued testing through driving simulators, which is particularly useful for rare “edge cases,” but this is no substitute for real-world robo-driver training.
Finally, general uncertainty about near-term travel activity may depress short-term interest in AV development. Will a large share of workers continue working from home? Will major cities that offer the most promise for many of the proposed AV business models see wealthier residents decamp to more bucolic telework settings? Will real or perceived health risks associated with shared transportation linger for an extended period of time? And these are just some of the known unknowns. These adverse trends may not come to pass, but the uncertainty will not be a boon to developers.
Despite all the challenges posed by Covid-19, the long-term outlook for automated vehicles remains positive. The large potential benefits to mobility and safety are unchanged. But we should accept that planned deployments may face further delays. One silver lining is that policymakers at various levels of government may have additional time to develop sensible regulatory frameworks that respect and promote AV innovation.