What We’re Reading
Waymo pitted its autonomous vehicles against a virtual superhuman driver to see which was safer (The Verge, September 29, 2022)
Spoiler alert: “the Waymo vehicle did better.”
In recent papers, Waymo presented the results of two studies it conducted to determine how a “virtual representation of a hyper-attentive driver”—who is “always attentive, never gets tired or distracted, and is always ready to react”—would perform against one of Waymo’s AVs in simulated tests. The research was intended to produce statistical support for the safety advantage of AVs over human drivers.
In the simulations, Waymo’s “superhuman” driver avoided 62.5% of crashes and reduced the risk of serious injury by 84%; the Waymo AV avoided 75% of crashes and reduced the risk of serious injury by 93%.
Waymo hopes that by publishing its peer-reviewed study results, “policymakers, researchers, and even other companies may begin to take on the task of developing a universal framework” for AV safety.
Growth in Autonomous Vehicles Could Improve Traffic Flow (Governing, October 2, 2022)
According to Carnegie Mellon researchers, AV adoption must reach 20% before we begin to see anticipated traffic benefits, but only 50% to realize most benefits.
These benefits “include the smoother, more choreographed flow of traffic that comes from having the vehicles talking to each other, making collaborative decisions around speed, routing and other factors.”
One impediment to reaching 20% adoption is the current regulatory landscape, which remains “structured around human-operated cars” and is not easily adaptable to disruptive technology like AVs. The National Association of City Transportation Officials, for example, has urged NHTSA and other regulators to “develop regulatory rules specific to automated driving systems (ADS), arguing they need their own regulatory environment.”
Tesla will remove more vehicle sensors amid Autopilot scrutiny (Reuters, October 4, 2022)
Tesla is moving forward with its commitment to develop camera-only autonomous driving, announcing that it will be removing ultrasonic sensors from its vehicles as of this month. Currently, Tesla vehicles have twelve ultrasonic sensors, located on the front and back bumpers. “The transition will temporarily limit automated parking features, but not affect crash safety ratings, Tesla noted.”
Experts remain skeptical that Tesla can achieve its goals without sensors. According to Professor Phil Koopman of Carnegie Mellon, “[t]he question is how well the cameras can see nearby the car, which sometimes can be limited.” Tesla hacker Green the Only tweeted that Autopilot relies on ultrasonic sensors as a “fail safe” for lane changes, but that “[i]mpact on safety of autopilot is likely small.”