Ready for a family outing to the mall and then dinner, you call your car. It backs out of the garage and waits in your driveway. All of you pile in and sit wherever you want, since no one will be driving. You face front while your spouse and kids swing their seats around to face each other.
You’ve told the car where to go, so it chooses the quickest route, obeying all stops and speed limits, keenly aware of what’s happening around it. Someone steps off the curb ahead and the car slows, ready to stop if necessary. Its front-facing motion sensors eye a dog cavorting on one side of the road. You’re catching up on emails, your spouse is texting, and the kids are playing video games. The car lets you out at the mall, then zips off to park. When your shopping is done, you call your car to pick you up and take you to your favorite restaurant. After dinner, you catch a few winks on your way home.
That’s the sort of scenario most people envision when they think of self-driving vehicles. And indeed, automakers and technology firms are investing billions of dollars in autonomous vehicle development on the presumption that AVs will be ideal for ride-sharing, ride-hailing, and deliveries — and that the driving public will ultimately embrace them for personal use, too, once they’re available and affordable. But considerable challenges remain, not the least of which is that many drivers actually enjoy driving and aren’t eager to relinquish control.
So how long will it be before we all start letting our cars drive themselves? Perhaps not as soon as some of us might hope and/or fear — nor as soon as safety advocates and auto execs would like.
Motor City Hard at Work
Here in Detroit and around the world, automakers are busy testing and developing AVs on both closed tracks and public roads while governments at all levels scramble to establish rules and regulations for safe AV operation.
General Motors’ Cruise LLC subsidiary has been testing Chevrolet Bolt electric-vehicle-based Cruise AVs in San Francisco and elsewhere while also, in a partnership with Honda, developing a fully autonomous — no driver, no controls — Origin AV for urban passenger and delivery service. Unveiled in January 2020, the self-driving, six-passenger Origin has production approval, and development prototypes are being tested at GM’s Milford Proving Grounds.
“We feel that Cruise has all the building blocks in place to lead in self-driving vehicles,” GM President Mark Reuss says. “And the first ones will be built right here at Factory ZERO, our Detroit-Hamtramck assembly facility. In October, Cruise received a permit from the California [Department of Motor Vehicles] to remove human backup drivers from its self-driving cars. That means Cruise can send its cars out onto the streets of San Francisco without anyone at the wheel.”
In fact, Cruise should have Bolt-based driverless AVs running around San Francisco by the time you read this. GM’s bold commitment to no-driver AVs, focusing first on city transportation, is one major element in its ambitious vision of a world with “zero crashes, zero emissions, and zero congestion.”
“This is our moonshot,” Cruise CEO Dan Ammann says. “The chaotic, gritty streets of San Francisco are our launchpad, and it’s where over 2 million miles of city testing will truly hit the road for the first time: an electric car, driving by itself, navigating one of the most difficult driving cities in the world.” In addition, Walmart plans to start testing automated deliveries using Cruise AVs in Scottsdale, Arizona, early this year.
Ford, in partnership with technology developer Argo AI, has been testing its own AVs (with safety drivers) on Michigan Avenue and around Michigan Central Station. The company has established AV terminals, command centers, and high-resolution mapping for ride-hailing and deliveries in Austin, Texas, Miami, and Washington, D.C., beginning in 2022, and is also testing in Pittsburgh and Palo Alto, California. The Ford/Argo AI team will assess the need for a safety driver and make a decision on when to operate without one based on the regulatory environment, safety performance data, and an appropriate level of community acceptance.
“We are very focused on Level 4 [see automation levels sidebar],” says Ford Autonomous Vehicle Technologies director John Rich, “removing the driver from the equation and operating autonomously within a geonet.” A geonet, he explains, is not the same as a geofence, which requires that the vehicle be able to drive anywhere within a metropolitan area. “Within a geonet, we will choose not to drive some places within that area, but our geonet will expand as we move forward,” Rich says.
Similarly, Cruise AVs “will not go into areas that we haven’t mapped,” Cruise Origin chief engineer Jason Fischer says. And while current Cruise AVs retain their steering wheel and pedals so a driver can take control if needed, the Cruise Origin does not. “There will be no ability to take control of the vehicle,” Fischer says. “The autonomous driving system will always be in control.”
Ford/Argo AI’s fourth-generation self-driving vehicles are Escape Hybrids equipped with the latest advanced sensing and computing technology. “We have upgraded our sensing suite with even more advanced lidar [like radar, but using laser light], higher-resolution cameras, and more capable radar sensors,” Ford Autonomous Vehicles chief engineer John Davis says. “Combined, this helps improve detection of fixed and moving objects on all sides … providing a blind-spot curtain, detecting things like a passing car or bicyclist in a nearby bike lane.”
FCA, meanwhile, has partnered with self-driving technology company Waymo. Launched in 2009 as the Google Self-Driving Car Project, Waymo has developed a Level 4 Waymo Driver system that powers Waymo One, a ride-hailing service, and Waymo Via for trucking and deliveries. The partnership claims to have logged more than 20 million miles of autonomous driving on public roads in 25 U.S. cities and 15 billion miles of simulation testing. It’s now offering AV rides to the public in Phoenix.
“Our now-four-year partnership with Waymo continues to break new ground,” FCA CEO Mike Manley says. “By incorporating the Waymo Driver — the world’s leading self-driving technology — into our Pacifica minivans, we became the only partnership actually deploying fully autonomous technology in the real world, on public roads.”
FCA is also testing Waymo Driver in Ram ProMaster vans for deliveries, and it plans to expand the technology across its product line.
Michigan has become a major hub for AV testing and development by multiple automakers and others. Two Michigan facilities — the American Center for Mobility next to Willow Run airport in Ypsilanti Township, and the 32-acre mock city called Mcity on the University of Michigan’s Ann Arbor campus — are dedicated to this ongoing work.
Meanwhile, an ambitious “connected corridor” linking downtown Detroit to Ann Arbor and Metro Airport along some 40 miles of Michigan Avenue and I-94 is in development. The initial vision calls for one dedicated interior lane for both the east and west sides of Michigan Avenue, which will need barriers at first to separate AVs from general traffic, including pedestrians. Crosswalks will be needed, traffic lights will need to be coordinated, and all manner of hardware and software will be required to connect GPS satellites, cellular arrays, Wi-Fi systems, sensors, and underground fiber cables.
The project will be managed by Cavnue, a subsidiary of New York-based Sidewalk Infrastructure Partners, working with Michigan’s Department of Transportation, Office of Future Mobility and Electrification, Economic Development Corp., and the Department of Labor and Economic Opportunity, along with state and local partners, stakeholders, and communities.
It will evolve to meet more transportation needs, but in the beginning, it will accommodate linked buses and shared mobility vehicles such as vans and shuttles. It will later expand to other connected and autonomous vehicles, such as freight and personal vehicles. Phase one is targeted for completion in the second half of 2022.
Roadblocks Ahead for Autonomous Vehicles
While self-driving vehicles hold the promise of eventually being much safer than those with humans at the wheel, significant challenges lie in their path to ubiquity.
“Automated vehicles’ potential to save lives and reduce injuries is rooted in one critical and tragic fact: 94 percent of serious crashes are due to human error,” the National Highway Traffic Safety Administration (NHTSA) contends in a report outlining the advantages of an automated-auto future. “Automated vehicles have the potential to remove human error from the crash equation, which will help protect drivers and passengers as well as bicyclists and pedestrians.”
One of the chief missions for AVs will be to provide much-needed mobility for the elderly and disabled, though ride-hailing services such as Uber and Lyft are already serving many Americans.
“Roads filled with automated vehicles could also cooperate to smooth traffic flow and reduce traffic congestion,” NHTSA contends. “With automated vehicles, the time and money spent commuting could be put to better use. … In many places across the country, employment or independent living rests on the ability to drive. Automated vehicles could extend that kind of freedom to millions more.”
To be as safe as envisioned, however, AVs will need to see, understand, analyze, and react to everything around them through a complex system of sensors, radar, and both visual and thermal cameras. All of that will add a lot of cost. And how effective will those systems be in darkness and nasty weather? When dirt covers their lenses? When snow blankets lane markers and road edges?
“Inclement weather is a challenge,” GM’s Fischer says. “We are working with suppliers on advanced cleaning systems that will help us solve those problems.” Adds Ford’s Rich, “All varieties of weather are being tested, and there will be a learning curve, with capability expansion over time.”
Will AVs be programmed to protect their occupants at the expense of others? Which way will they dodge if they can’t stop to avoid a pedestrian when the alternative may be an oncoming vehicle, a tree, a lake, or a cliff?
Most recorded AV accidents so far have been minor, often when other vehicles hit the AV. But the few fatalities have been widely reported: An Uber test AV hit and killed a pedestrian in Arizona when its “safety” (human) driver was distracted. A Tesla on “Autopilot” killed its driver when it drove under a semi-trailer that it didn’t detect. Another Autopiloted Tesla killed a passenger, and a third killed two people in another car in other incidents.
Tesla CEO Elon Musk contends self-driven Teslas are safer than those under human control but that their Autopilot can enable extended lapses of driver attention. As of September, NHTSA had investigated 23 crashes involving vehicles using “automated assistance” systems, 14 of them Teslas.
Such incidents, though rare, inevitably reduce peoples’ trust in self-driving vehicles, and recent studies show that only about half of Americans say they are willing to ride in one.
“We have to make these vehicles better than humans,” Rich says, “constantly alert with better reflexes and better ability to avoid an accident. They may never be perfect, but if they are considerably better than humans, we almost have a moral imperative to put them on the road because we will be saving lives.”
Another fear is that AVs will be rolling roadblocks, obeying all posted speed limits while everyone else swerves around them? “The vehicles are programmed to obey the law,” Rich points out. “We won’t be able to speed or do a lot of things you see human drivers doing today.”
And when someone inevitably is hurt or killed despite everyone’s best efforts, who will be liable? The vehicle’s owner? The manufacturer? The software programmer? The town or city where the incident occurs? All of the above? “Initially, the lawyers will sue everyone involved,” says Carla Bailo, CEO of the Center for Automotive Research (CAR) in Ann Arbor. “As these cases are settled and precedent established, it will become more clear.”
The State of the Art
While that family-outing scenario of Level 5, aka “Full Automation,” for privately owned vehicles looks to be a long way off — if it ever arrives at all — “Level 4 is essentially here now,” CAR’s Bailo says. “Level 5 is later, pending many other nontechnical parameters such as regulation, public policy, legal, and insurance.”
No current AV is intended for private ownership. “They will be able to move goods and people in a controlled environment,” Rich says, “but you will not be able to go out and buy one. They are difficult to manage and will require professional service to run.”
Adds veteran auto journalist and AutolineTV host John McElroy: “Autonomous cars that can drive anywhere and that you can buy at a dealership will not be available this decade. Maybe next decade. Tesla claims otherwise with its Full Self Driving, but it’s up to them to prove it, since it’s been delayed multiple times. That said, 2020 [was] really the year of autonomous vehicles. They’re on the streets and running now. The technology is available, and it works. It’s expensive, but the cost is coming down fast. For now, AVs are relegated to geofenced areas that have been 3D-mapped, but those fence posts keep moving.”
The good news is that Level 2 , aka “Partial Automation,” is available to the driving public today. Many new vehicles, even at affordable prices, offer Adaptive Cruise Control, which adjusts speed to maintain a set gap behind the vehicle ahead, and Lane Keeping Assist, which keeps your vehicle in its lane. That combo allows hands-off cruising for a few seconds in scenarios where road edges and lane markers are clearly visible to the vehicles’ cameras. Some systems work better than others, you must pay full attention and be ready to take control at any time, and the systems tell you when to take the wheel. If you don’t, they shut off.
One of the best available systems is Cadillac’s Super Cruise (available on some models now and being expanded to more), which will soon add a feature by which it can change lanes itself. GM says its ultimate Super Cruise vision is to offer hands-off driving capability 95 percent of the time when the vehicle is on “enabled” — that is, precisely GPS-mapped — roads.
While we may be a long way from a world where all vehicles drive themselves, the not-too-distant future may well bring a potentially worrisome mix of driverless AVs sharing the roads with a large majority of human-driven cars and trucks. The AVs will be capable of communicating vehicle-to-vehicle and vehicle-to-infrastructure to avoid conflicts with each other but will still have to continuously monitor everything around them and make assumptions (as alert drivers do) about other vehicles’ expected behavior.
Will you trust a vehicle with no driver or controls to shuttle you around, or will you prefer a human-driven Uber, Lyft, or taxi? Or will you continue piloting those trips behind your own wheel for as long as you can? If you’re not yet AV-ready, perhaps you will be when your own capabilities someday diminish.
Levels of Autonomy
The National Highway Traffic Safety Administration (NHTSA) foresees a future in which fully autonomous vehicles are integrated onto U.S. roads and highways. To get to that point, though, they will have to progress through six levels of technological advancement — from no automation at all to full autonomy. Here’s how the NHTSA defines those six levels:
No Automation. Zero autonomy. The human driver does all the driving.
Driver Assistance. The vehicle is controlled by the driver, but an advanced driver assistant system (ADAS) in the vehicle can sometimes assist the human driver with either steering or braking/accelerating, but not both simultaneously.
Partial Automation. An ADAS in the vehicle can control both steering and braking/accelerating simultaneously under some circumstances, but the human driver must continue to pay full attention at all times and perform the rest of the driving task.
Conditional Automation. The human driver is necessary but not required to monitor the environment. An automated driving system (ADS) in the vehicle can perform all aspects of the driving task under some circumstances, but the driver must be ready to regain control at any time at the ADS’ request. In all other circumstances, the human driver performs the driving task.
High Automation. An ADS in the vehicle can perform all driving tasks and monitor the driving environment — essentially do all the driving — in certain circumstances. Human occupants need not pay attention in those circumstances but may have the option to control the vehicle.
Full Automation. An ADS on the vehicle can do all the driving in all circumstances. The human occupants are just passengers and need never be involved in driving but may have the option to control the vehicle.