What Barak Gila's roughly 15-minute have data), but they're killed at higher rates commute through San Francisco's Castro than vehicle occupants. Aside from a slight and Mission Districts lacks in length, it dip in 2020 when we drove less early in the makes up for in excitement. Gila, a software pandemic, pedestrian and cyclist fatalities engineer, knows that on any given day his have risen for over a decade. In 2021 the route will be dense with traffic—some annual total of cyclists killed jumped 5 pedestrians and bicycles, but mostly motor percent to an all-time high of nearly 1,000, vehicles. Those come in every size and according to preliminary data from the permutation possible, including the National Highway Traffic Safety self-driving cars (also called autonomous Administration (NHTSA). Pedestrian vehicles, or AVs) that developers like fatalities were up 12.5 percent to 7,388, the Google's sister company Waymo and the highest since 1981. General Motors-owned Cruise are testing in the city. But an autonomous vehicle will never be distracted by a text message; nor will it Gila commutes by bike, and doesn't mind drink and drive or road-rage, says Anne riding around AVs. While vehicles from Dorsey, a software engineer in Waymo's both Waymo and Cruise have been behavior division. Removing humans from documented exhibiting alarming behavior in the driving task, or dramatically reducing the city—swarming, blocking traffic, even their role, could save thousands of lives and rolling through an active firefighting countless injuries every year, especially scene—Gila says he's "never had a among vulnerable road users. self-driving car behave unsafely" around him. And it wasn't an autonomous vehicle That's the promise, anyway. But autonomy's that right-hooked him in May 2021; it was a safety benefits aren't yet proven, and even if human-driven Porsche, whose driver told they do pan out, a decade of halting him the car's blind-spot detector hadn't development progress suggests that rolling alerted him to Gila's presence. Gila had been out true self-driving vehicles on a scale that vigilant and was able to avoid injury, but the could achieve those gains will likely take far episode was a troubling harbinger. longer than what the AV industry has promised. In the meantime, car An article of faith among proponents of manufacturers are pushing forward with autonomous vehicles is that most (94 advanced driver assistance (ADAS) percent is often cited) traffic crashes are technology, offering "autonomy lite" caused by human error. Pedestrians and features like the ones in that Porsche. But people on bikes made up a minority of the that approach comes with its own issues: 42,939 road deaths in the United States in Studies suggest that the misleading marketing and tech terminology is often so Cyclists often don't know if someone will confusing that a frightening number of wave them through even when they have the drivers treat their cars as self-driving when right of way, or ignore them and cut them they're not. off. In the same scenario with a self-driving vehicle, the car should react the same way As Gila found out, that technology—which every time—and not only that car, but every ranges from blind-spot detection to systems car in the fleet. The cars share the same that can handle all driving tasks in limited software, so it's essentially as if they all have conditions—is far from foolproof. "The the same driver. This fact makes safety [driver] was relying on it, but it doesn't work scalable, says Clay Kunz, a robotics 100 percent of the time," Gila says. "That software engineer in Waymo's perception false promise of safety can be almost worse division. Say a car responds unsafely to a than nothing." pedestrian during testing. A software update can change the behavior of every vehicle In terms of perception abilities, the Level 4 with that operating system. "Once you solve autonomous vehicles that Waymo and others the problem," he says, "the behavior is are testing are already vastly more capable distributed across the fleet." than human drivers. (According to the Society of Automotive Engineers, Level 4, But proper behavior—which involves or "high automation," means the vehicle judging what other road users are doing and drives itself while the occupants are then responding appropriately—is hands-off passengers. But it can operate exceptionally difficult because of the only in limited service areas; full-on, massive variety of possible driving go-anywhere autonomy would be Level 5, scenarios. "Cyclists and pedestrians are which is almost a sci-fi dream.) Level 4 challenging because they can be anywhere vehicles use a sophisticated array of sensors, in the driving space," says Stephanie including high-definition cameras, Villegas, Waymo's former lead of structured microphones, radar, and a kind of powerful testing. They can be following traffic rules laser scanner called LiDAR (light detection or going against traffic. They can be in a and ranging) that can create lane or between lanes, and the difference three-dimensional maps of the driving matters a lot insofar as predicting what environment. they'll do next. "It's hard to list all the ways they interact with drivers," she says. Another area where autonomous vehicles That's a challenge even for human drivers, have an edge over humans is consistent but we have a built-in advantage. "Humans behavior. "If the Driver responds the same are really good at predicting the intent of way every time, that can increase safety for other humans based on things like posture cyclists because they know what to expect," and explicit gestures," says Justin Owens, says Dorsey. Consider a four-way stop with PhD, a research scientist with the Virginia human drivers and someone on a bicycle: Tech Transportation Institute. "We're hardwired to do that, with capabilities we've evolved over millions of years." As Sam Anthony, founder and former CTO of the autonomous vehicle software company Perceptive Automata, wrote in a 2022 Substack post, it takes a fraction of a second for a human driver to see a pedestrian and process a massive amount of contextual information on the person's age, attention level, and even emotional state, all of which influences how the driver responds.
Even if autonomy proves its safety
argument, the obstacles of scale render its benefits unattainable in the near term. Right now, Level 4 autonomous driving exists only in small fleets; Waymo operates 300 to 400 vehicles in Phoenix, for instance, and just 700 total. Even if companies like Waymo and Cruise succeed with their autonomous taxi business, they may not account for a large enough slice of traffic—at least 20 percent by one recent study—to create noticeable safety improvements. If, tomorrow, every new car sold were fully autonomous, it would be almost four years before even 20 percent of the 276 million registered vehicles in the U.S. were self-driving. Of course, that's a totally unrealistic scenario; the reality is, any safety benefits from autonomy are almost certainly a decade or more away.
Autonomous vehicles, or self-driving cars, are innovative technological advancements that can rapidly discern their circumstances and function without human interference, using sensors, cameras, and GPS technology to direct the roads