Paving the way for a smarter, safer future of autonomous vehicles

In April 2022, a fully autonomous all-electric Jaguar I-PACE picked up a Waymo engineer in San Francisco, stopped to pick up a coffee, and then took the engineer to work. No one was in the driver’s seat, just a steering wheel turning as if by a ghost.

Waymo, an autonomous driving technology company, has been shuttling passengers to destinations in the Phoenix and surrounding suburbs since 2017; the company launched a fully driverless service there in 2020. Waymo is an emerging leader in the autonomous vehicle (AV) space, having completed more than 100,000 rides since 2017. Its introduction of driverless taxis on the steep, densely packed streets of San Francisco seems like a breakthrough for the technology, which has been pummeled in the press over safety concerns and chided by naysayers.

Waymo is not the only company eager to get commercially viable AVs on the road. According to a report from CNBC, VW-backed Argo, which uses an attending driver, is being hailed by Lyft passengers in Miami Beach, FL. US companies such as Tesla, Zoox, and GM’s Cruise subsidiary have raised tens of billions of dollars to test self-driving algorithms and sensors on roadways spanning thousands of miles.

Today, there are 1,400 self-driving cars in the United States being tested by more than 80 companies. In Beijing, Pony.ai and Baidu have deployed robotaxis in a fast-growing market: Globally, self-driving cars are expected to account for one in 10 vehicles by 2030 and generate 1.2 trillion dollars from robotaxis alone. Today, the global AV market sits at $54 billion.

“The benefits are [AVs] enable movement without a driver’s license, and they ultimately will produce greater safety,” says Robin Chase, co-founder and former CEO of Zipcar and founder of the New Urban Mobility alliance (NUMO), an urban mobility nonprofit based in Washington, DC.

“Today, we as a society have accepted crazy fatality and accident rates,” Chase says. “And that will not be the case for autonomous vehicles.” Additionally, AVs could expand mobility “to people who are blind, people who have epilepsy, or people who are too young or too old.”

AVs could also reduce the number of cars on the road, freeing up space used for on-street and sheltered parking. In the United States, about 75.9% of people commute to work alone. As part of fee-based public transit systems with carbon incentives like congestion pricing, Chase says, automated rideshares and shuttle fleets could significantly decrease the 6.9 billion hours Americans spent in traffic in 2014 and reduce private car use. A University of Texas study found that if 10% of the cars on US roads were autonomous, approximately $25.3 billion could be saved in time and fuel expenditures, while also reducing injuries and deaths.

Israel Duanis, head of autonomous vehicles at mobility software company Via, says 70% of RAPID riders—an on-demand AV car service in Arlington, TX, that uses sensing technology from May Mobility—use the system for essential destinations, including work, school, and health care. Since its launch in March 2021, the automated service has conducted 30,000 rides; 80% of riders reported feeling safe when riding RAPID.

Currently, 94% of the 35,000 fatal crashes on US roads each year (about 12 deaths per 100,000 people) are caused by human error. The Insurance Institute for Highway Safety reports that the current car fatality rate is 1.34 deaths per 100 million miles traveled. Some experts estimate that rate could decrease by a factor of 10 by removing accidents caused by drunk drivers and texters—another metric in favor of AVs.

To “see” the road, autonomous cars rely on cameras, radar, and LiDAR (laser radar). By bouncing a laser off an object and measuring the return time, LiDAR can judge distances and “see” in 360 degrees. If high-resolution sensing tools are the eyes of self-driving cars, their cognitive abilities come from intelligent algorithms and artificial intelligence that interpret this raw data against reference maps, says Sravan Puttagunta, former CEO of California-based Civil Maps.

According to Puttagunta, the Civil Maps platform (like those of competitors Mobileye, Delphi, and Bright Box) gives an autonomous vehicle contextual awareness, not just a 2D navigation map. It uses artificial intelligence to process raw data coming from high-resolution laser imaging, so cars can localize precisely and make better tactical decisions—like what to do at a four-way stop or roundabout.

“What we’re working on is [the problem of AVs having an] abundance of caution,” says Paul Newman, founder and chief technology officer at the Oxfordshire-based driverless vehicle software company Oxbotica. “How is an AV going to be appropriately assertive at a rotary? Because the worst thing you can be is unknown, unsafe.

“There’s all kinds of noise and religion about the right algorithm you should use,” Newman adds. “I think that’s all for the birds. It’s an architectural systems problem.”

In other words, AV safety is a matter of redundancy: equipping vehicles with overlapping sensors and algorithms to avoid system gaps that increase accident risks and stall travel. When human drivers see a swerving car or hear a wailing siren, they tend to slow down. AVs need similar foresight, Newman says, to anticipate dangerous situations and pull to the side of the road when needed.

AV technology has advanced dramatically since Japan’s Tsukuba Mechanical Engineering Laboratory launched the first semiautomated car in 1977. Commercial vehicles like forklifts and mining vehicles have been operating autonomously for decades, Newman says.

Now, autonomous last-mile delivery vehicles are cropping up in cities such as Houston and Mountain View, CA. Nuro’s zero-emission refrigerated mini-cars are backed by brands such as Kroger, 7-Eleven, Dominos, Walmart, FedEx, and Chipotle. They operate on renewable energy from wind farms in Texas and carry as much as 500 pounds of groceries, according to a company spokesperson. These small-scale urban mobility vehicles could be integrated into broader supply-chain networks in which electric autonomous fleets, running nearly continuously, could transport goods faster, more safely, and with less fuel.

“The first use cases are going to be places where we can use the vehicle as many hours per day as possible, because they’re expensive,” Chase says. “So that means taxi-type services, delivery, or trucking.”

Aurora, which acquired Uber’s self-driving division in 2020, is pilot testing its AV software with Werner Enterprise tractor-trailers on a 600-mile span of highway in Texas; FedEx and Uber Freight have run similar programs elsewhere. Scaled across long, monotonous stretches of highway, similar AV platoons could narrow a gap in the nationwide shortage of qualified human drivers.

It could be years or decades before self-driving cars become a mainstay of cities and highways, says David Zuby, chief research officer at the Insurance Institute for Highway Safety. Ensuring safe coexistence with other vehicles is a key hurdle. Fully autonomous vehicles, at least for now, must share roads with cars operated by human drivers and those with conditional autonomy—both present safety risks.

The National Highway Safety Administration defines levels of driver-assistance technology from zero to five. At Level 0, the driver is fully responsible for the vehicle, and must remain “engaged and attentive” at all times. At Level 5, “occupants act only as passengers and do not need to be engaged.” Most private and commercial vehicles operate somewhere in the middle: Level 1 to Level 3.

Chase says that semiautonomous cars—particularly vehicles with automated functions like lane centering, acceleration, and braking controls for highway driving (Level 2) or cars that require a safety driver to be alert and engaged when things go wrong (Level 3)—can be more dangerous than manually operated vehicles because they can lull drivers into false complacency or drowsiness.

A 2016 accident, in which 40-year-old Joshua Brown died while driving a semiautonomous Tesla Model S, tragically illustrates this complexity. A truck made a legal left in front of Brown’s Tesla, but instead of stopping, the car sped under the truck’s trailer. According to Mike Demler, a senior technology analyst at The Linley Group, the Tesla’s detection system of sensors, radar, GPS, and image-processing software was not intended to be used in hands-free mode. “The sensors weren’t designed to account for cross traffic, only highway driving,” Demler says. “They could identify the backside of a car, not the side of a tractor trailer.”

Of the nearly 400 crashes of partially autonomous vehicles reported in between July 1, 2021, and May 15, 2022, roughly 70% (273 crashes) involved Teslas using Autopilot, "Full Self-Driving," Traffic Aware Cruise Control, or other driver-assist systems. These results are not necessarily a categorical indictment of Tesla, which is one of the most prevalent manufacturers or semiautonomous vehicles with 830,000 vehicles on the road—but it is emblematic of the safety risks they pose, Chase says.

“If on 10,000 of your trips you don’t have to pay attention, everything’s fine, and then on the 10,001st trip, you’re responsible—you’re supposed to be paying attention and take control—I think that’s an impossible situation,” Chase says.

The US Department of Transportation is working to modify existing vehicle safety regulations to better accommodate situations with no human driver. Questions around how to assign liability for crashes involving cars operating autonomously or in self-driving mode is a murky legal area that Zuby says must be clarified within the next several years, as Level 2 and Level 3 cars become available at lower costs.

Aside from regulatory issues, Zuby says US roadway infrastructure and laws are largely determined by individual states and present varied speed limits, highway striping, and red-light positions at intersections.

“The companies developing automated driving systems have to deal with the infrastructure that’s there,” Zuby says. “Certainly, [it would be advantageous] if we could get a coordinated effort to make improvements that mediate the development of automated driving systems. But I think it’s a long shot that somehow all 50 states will get together and agree to make the roads more uniform for automated vehicles.”

European and Asian cities may be in a better position to accommodate a transition to AVs, given their focus on effective public transit systems—especially in fixed geographies with intensively used mass-transit routes. “European cities are increasingly working to pull private cars out of the cities or diminish the use of private vehicles,” Chase says. “The US has yet to make that a serious commitment or even a stated assumption.”

The future of public transit, she adds, is “not going to be all AVs. It’s going to be AVs and bicycling and walking and public transit—resilient, multimodal ways to get around.”

In the absence of overhauling US cities’ existing infrastructure and mass-transit systems, software logistics companies such as Via (along with competitors like Uber and Lyft) may play an important role in AV development. These companies are positioned to coordinate rider journeys across connected multimodal transit systems, with intelligent on-demand booking, routing, and dispatching services.

In addition to providing logistics services for Arlington RAPID, for instance, Via supports the A2GO Lexus RX 450h shuttle service in Ann Arbor, MI. Both services are available to the public and designed to be part of the existing mass-transit infrastructure. In February 2022, Via launched an AV service with Motional in downtown Las Vegas focused on serving essential travel for residents of the city. Services in Kelheim, Germany, and Minnesota that will demonstrate the impact of AVs in rural environments are also on the horizon.

“Communities will soon move beyond short-term pilots with small fleet sizes [to create] permanent AV networks that are fully connected within a community’s existing transit ecosystem,” Duanis says.

The discussion around AVs necessarily becomes a discussion on policy and infrastructure. A 2018 study from UC Berkeley estimated a fleet of shared automated electronic vehicles operating in Manhattan with a battery range of 50–90 miles could save $0.29–$0.61 per passenger mile over current taxi services—while reducing greenhouse-gas emissions by 73% and energy consumption by 58%. The program would require a $10 million capital investment in 1,500 charging stations (equivalent to roughly $0.01 per passenger mile).

“Our results were surprising because they contradicted current market trends pushing toward long battery range and superfast charging,” says Via Data Science Principal Gordon Bauer, one of the study’s authors. “Whereas in the private vehicle market, drivers want something that will work for multiday road trips, Manhattan taxi trips are typically very short, with downtime in between when vehicles could be charged.”

Six years after the study, it’s still unclear when, if ever, AVs could serve as robotaxis in Manhattan. But many of the conclusions also apply to human-driven electric-vehicle fleets. In subsequent studies, Bauer analyzed transportation networks in New York, San Francisco, and New Delhi, and found similar results: Many fleet services could be electrified immediately, saving money and reducing carbon emissions.

The pending launch of Oxbotica’s autonomous mobility software in electrified 14–16 person shuttles could be a bellwether for future AV adoption. Oxbotica, in partnership with the German automotive company ZF, plans to roll out shuttles in the foreseeable future. Newman says the vehicles will handle all driving functions over limited urban areas, requiring remote human assistance only to resume routes after stopping for exceptional circumstances, such as a passing ambulance or a flooded street.

Newman sees such electrified transit services unleashing a cascade of benefits, analogous to when wolves were reintroduced at Yellowstone National Park: “When wolves were introduced, they set off a string of ecological benefits: Elk packs became healthier. Fewer elk meant less overgrazing, which led to the return of native plants and pollinators. That led to less stream runoff, which brought back fish and beavers. Beavers built dams that slowed the rivers and changed their shape.”

If he’s right, this may be the cusp of a safer, cheaper, more fuel-efficient way to travel—but only time will tell.

This article has been updated. It originally published in March 2017.