Robotaxis: Should we be on board?

Robotaxi programs are on the rise in the United States. 

Since 2018, when Waymo debuted the nation’s first commercial robotaxi service in Arizona, driverless rideshare services have launched in San Francisco and Los Angeles. The company has also begun testing in Austin, Texas, and plans to introduce service in Atlanta in 2025. Last spring, New York City began selling permits for robotaxi companies to test their services, with safety drivers in the cars to correct course if needed. 

But even as Waymo and its competitors steadily expand operations, several high-profile incidents have stirred concern about robotaxis and other autonomous vehicles.  

Some of these incidents have garnered attention for their absurdity, such as when an entire lot of Waymo robotaxis in San Francisco honked at each other through the night. But others have had grave consequences.  

In October 2023, a San Francisco pedestrian struck by a conventional car landed in the path of a Cruise vehicle, which dragged her twenty feet. The accident led Cruise to pause its robotaxi program for several months before resuming mapping operations in Phoenix the following April. By December 2024, parent company General Motors announced that it would cease funding the Cruise program.

So, are robotaxis and other autonomous vehicles here to stay? Here’s what you need to know: 

What is an “autonomous” taxi, anyway? 

It depends on how you define the term. The vehicle industry group SAE International categorizes autonomous vehicles on a scale ranging from 0 to 5. Level 0 represents cars with only the most basic driver support, such as blind spot warnings. Vehicles that perform all driving tasks under all conditions are classified as SAE Level 5, but so far, no manufacturer has produced a car with this level of autonomy. 

All robotaxis are currently classed as SAE Level 4, meaning they can operate without a driver most of the time but require human input in some circumstances, which is why they need a backup: either an in-person safety driver or a remote operator in a control center who is notified when a problem arises.

How do robotaxis work? 

Like other autonomous vehicles, most robotaxis rely on several types of sensors to gather information about the outside world, drawing from a combination of cameras, ultrasound, radar, and lidar, a technique that measures distances using lasers.  

The data the car collects is interpreted by the vehicle’s software, which makes decisions about what actions to take. If the system signals that it does not know what to do, and there is no safety driver on board, a remote operator needs to be ready to step in quickly. In a 2024 blog post, Waymo said remote operators receive questions from the autonomous system and provide information about how to proceed.

Is a remote operator really a safe option if something goes wrong in a robotaxi? 

In-person safety drivers in California are required to pass a background check to operate a robotaxi, but no such regulations are in place for remote operators, according to Phil Koopman who studies autonomous vehicle safety at Carnegie Mellon University. Remote assistants are “not regulated as drivers, but they can cause crashes,” he says. 

Koopman says there is no federal law requiring remote operators to even have a driver’s license. Companies may have their own policies in place to ensure remote operators are up to the task, but the autonomy companies have in selecting remote operators is a major concern to some experts. 

“You could be drunk or high and remotely operating a vehicle,” says Missy Cummings of George Mason University, who previously served as a senior safety advisor at the National Highway Traffic Safety Administration. She hopes federal policymakers will write rules to ensure remote operators can do the job safely.

Koopman worries that one day, companies could choose to outsource remote assistance to other countries, which would make oversight and law enforcement more difficult. 

How are robotaxis tested for safety? 

Autonomous vehicle technology testing is regulated by the National Highway Traffic Safety Administration, but the decision on whether a robotaxi company can begin operating in a particular community is up to state regulators. California’s rules are the strictest, according to Cummings, who says the state’s regulations could help predict eventual federal guidelines. 

Robotaxis are tested on a city-by-city basis. In California, regulators require in-person safety drivers as backups during the initial testing phase in a community. Even if a robotaxi service performs well with in-person safety drivers as backups, there’s still the question of whether the safety drivers can be safely phased out in favor of remote operators.  

“The companies themselves decide when it is time to remove the safety driver, and we have seen some companies get that decision wrong,” Koopman says, citing the 2023 Cruise pedestrian dragging incident as one example of when safety drivers were removed prematurely. “Even in California, where they’re regulated, the regulators do not have enough information to know whether the safety drivers got pulled out too early or on time.” 

Cruise vehicles only operated with in-person safety drivers as backups during the period between the 2023 accident and the program’s anticipated end in early 2025. Prior to the announcement that their service would shut down, a representative for the company said in an email that Cruise had implemented recommendations from a review concerning the incident, including a software update to the collision detection system.

The testing process for driverless cars may change during the second Trump administration. Trump ally and Tesla CEO Elon Musk plans to begin producing Cybercab robotaxis by 2026 and has expressed interest in a single federal regulatory framework to replace individual state regulation of autonomous vehicles. According to Reuters, Musk could influence Trump in choosing the next Secretary of Transportation, who oversees the National Highway Traffic Safety Administration.

What other concerns have come up with robotaxis? 

Autonomous vehicles like robotaxis could disrupt emergency services. A patient died in September 2023 after a stalled Cruise vehicle blocked an ambulance on its way to the hospital, according to eyewitness accounts from the San Francisco Fire Department.  

A Cruise spokesperson disputed the fire department’s characterization in a statement to The San Francisco Standard. In a blog post the following month, the company announced that first responders would be able to unlock and move stalled Cruise vehicles in emergency scenarios. 

Robotaxis have also created traffic flow problems during major events. During a presidential campaign stop in San Francisco, Vice President Kamala Harris’ motorcade was blocked by two Waymo cars.

What about regular traffic congestion? Will robotaxis make it better or worse? 

While robocabs could be a convenient option for some consumers, data scientist and economist Bryan Weber of The College of Staten Island CUNY doubts they will reduce traffic problems. “I’m not horribly optimistic,” he says. 

Weber predicts that robotaxis, like traditional ridesharing services such as Uber and Lyft, will ultimately result in more cars on the road. When an empty robotaxi leaves a drop-off point to pick up the next customer, it is contributing to traffic without contributing to mobility. 

Self-driving cars may also not maneuver as efficiently as human drivers. Weber describes autonomous vehicles as “very conservative with their driving choices,” as they lack a human driver’s intuition about traffic patterns. 

This cautious programming is designed to reduce traffic accidents, and one analysis of car accident data by University of Central Florida researchers suggests that SAE Level 4 vehicles do behave differently than human drivers. Autonomous vehicles proved less accident-prone than human drivers in foggy conditions, but performed worse than human drivers during dawn and dusk hours.

While safety on the road is paramount, Weber expects that these inefficient choices will still add up, increasing congestion on the road.

What have we learned from robotaxi tests so far? 

Cummings says the results of robotaxi testing programs have varied by community, which she attributes in part to the different landmarks and forms of transportation in each city. For example, a Cruise car bumped into an articulated (“bendy”) San Francisco bus in March 2023 because it did not predict how the second half of the bus would move, according to a statement from Cruise. 

Waymo successfully introduced its technology in Arizona, Cummings says, but was less fortunate in San Francisco, where its fleet has been linked to incidents including a February collision with a cyclist that caused minor injuries. 

A Waymo vehicle struck and killed a dog in San Francisco in a 2023 accident that the company characterized as “unavoidable” in a statement to TechCrunch. Animal-vehicle collisions are not unique to autonomous vehicles; State Farm estimates that nearly 2 million claims related to such crashes were filed in the past year.  

Rising tensions between San Francisco residents and robotaxi companies last year led protestors to disrupt operations by placing traffic cones on top of vehicles, which confuses their operating system. In February, a crowd threw fireworks into a Waymo robotaxi, setting it on fire. 

However, some San Franciscans may have changed their tune. In late October, reporter Meghan Bobrowsky of The Wall Street Journal observed that more of the city’s residents have embraced Waymo in recent months as the program has opened to the public, allowing those who may have been skeptical to try the service. She noted that additional months of testing and competitor Cruise’s period of withdrawal from the city could have also played a role. 

Can we fix the problems with robotaxis?  

Autonomous vehicle technology isn’t likely to keep advancing as quickly as it has in recent years, according to Weber. “I think we’re past the point of leaps and bounds improvements” as further big innovations will be too expensive, he says. Still, Weber thinks that even marginal improvements will help. 

It’s not just about improving technology in the vehicles. More attention to the role of humans could also help improve robotaxi safety. While Koopman believes that in-vehicle safety drivers are the safest choice, he says the safety of remote operator backups can be improved with greater oversight. For example, federal guidelines in China require remote operators to undergo training. Remote operators at Chinese companies like Baidu assist vehicles using a steering wheel rather than a mouse and keyboard, as is the practice among remote operators in the US. 

Limiting the scope of service could also reduce the navigation problems robotaxis have faced so far. Cummings praises Aptiv’s now-defunct Las Vegas service that shuttled passengers between the airport and the Strip as a “great use” of autonomous vehicle technology because they operated at slow speeds and could reduce the amount of cars on the road.

“As long as you keep things slow, there’s a good chance that you could have effective operations,” Cummings says. 

Editor’s Note: This article was updated 11:30 p.m.Dec. 11 to reflect General Motors’ Dec. 10 announcement of the end of its Cruise program in early 2025.