The Washington PostDemocracy Dies in Darkness

Autonomous shuttles in Northern Virginia suburb show why the future of robot cars might be slow

At the site of Reston’s new Halley Rise development, Optimus Ride’s automated shuttles transport office workers whose routines have been disrupted by construction. (Jahi Chikwendiu/The Washington Post)
Placeholder while article actions load

At a building site in Reston, there is something small and quiet on the roads alongside big noisy dump trucks: automated shuttles that ferry office workers whose routines have been disrupted by the construction.

The shuttles’ job is to make runs from parking lots that have been shifted farther from people’s offices. But when the new offices, apartments and stores go up on the 36-acre Halley Rise development in Northern Virginia and a Metro station opens nearby, the company that operates them hopes the vehicles will speed people on the first or last mile of their daily commute.

The project, a partnership between Boston-based start-up Optimus Ride and real estate developer Brookfield Properties, is a glimpse of one version of a self-driving future that features squadrons of robotic vehicles that are smaller and slower than traditional cars and that can cart people efficiently over small distances.

Operating at low speeds — typically under 25 mph — in the confines of a private development is a way of sidestepping some of the technical and legal barriers that apply on the open road. And for a start-up such as Optimus, it offers the chance to compete with big tech firms and automakers. The company says it has completed some 15,000 rides in Reston since launching a few weeks ago.

Self-driving cars reach a fork in the road, and automakers take different routes

“If you do it here you can validate business models, you can validate the technology,” said Ryan Chin, Optimus Ride’s chief executive and one of its founders. “You can introduce it in a very safe way. You can get paying customers.”

Interest in low-speed autonomous vehicles has exploded in the past year, especially after a self-driving Uber SUV struck and killed a woman last year in Arizona, dampening enthusiasm around autonomous cars.

Kelley Coyner, the author of a forthcoming study, said she is aware of about 225 projects or proposals in the United States and a couple dozen more in Canada.

Optimus, a spinoff from the Massachusetts Institute of Technology, launched a shuttle service in Brooklyn over the summer and has at least one major direct competitor.

Traveling under 25 mph means there is less risk of killing someone if a pedestrian is hit, and the vehicle requires less-sophisticated sensors because stopping distances are shorter, said Sam Abuelsamid, an analyst at Navigant Research. But even for slow-speed vehicles, there are still significant technical hurdles to overcome.

Huei Peng director of Mcity, an autonomous-vehicle-research center at the University of Michigan, said the technology is advancing, but even low-speed self-driving cars have severe limitations. He compared them to the Wright brothers’ early airplanes.

‘Raising the bar’: Uber details shortcomings in self-driving car that killed pedestrian

“They flew a very short distance: not very high, not very far, not very fast,” Peng said. “They were not very exciting. They were not very useful.”

Perhaps the most obvious challenge: getting the vehicles to be truly driverless. Optimus uses modified six-seat electric buggies manufactured by Polaris, but two of the seats are occupied by a safety driver and an engineer.

Chin is confident that his company has solved the fundamental problem of automated driving and said that deploying the vehicles in the real world with crew on board is an important step toward going fully driverless, providing the chance to collect more data. He said the company’s goal is to switch from onboard crew to a remote operator managing a squadron of vehicles at one of its sites by next year.

Seemingly simple problems must be thought through: The shuttles in Reston don’t have sensors to alert them if someone opens the door while the vehicles are in motion. How should a vehicle handle a parent loading a child into a stroller? And will vulnerable people feel safe riding alone with strangers?

Andy Alden, a researcher at the Virginia Tech Transportation Institute, said the autonomous-vehicle industry was beset by “excessive optimism” as the navigation technology got better, only to be confronted by the people who use the vehicles or share the roads with them.

“There’s so many real-life issues,” he said.

So, for now, the crew remain on the Optimus shuttles. In Reston on Monday, John Sgueglia sat behind the steering wheel of one of the vehicles, just in case. Engineer Ruijie He sat next to him, a laptop open with its screen showing the world as the shuttle saw it — blobs representing the trucks and people its cameras and lidar sensor could see — as it completed a circuit from parking lot to office building.

California launches system allowing driverless cars to ditch their backup drivers

At times, the ride was smooth and it was difficult to tell that no human was in control. The shuttle deftly swerved between a pedestrian and an oncoming vehicle. Other times, the vehicle’s robotic brain appeared confused, lingering at an all-way stop and juddering when a group of pedestrians crossed in front.

From the front seat, He said some of that unusual behavior is purposeful. Inching forward sends a signal regarding where the vehicle is aiming to go, and behaving conservatively at stop signs helps protect against drivers who are in a rush.

Engineers working on self-driving cars have come to realize that there is a bewildering array of situations they need to be able to accommodate on the road. Human brains are pretty good at adapting to novel situations; robots, not so much.

Chin said driving in a parking lot can be quite challenging — demonstrated in recent online videos of electric Teslas struggling with a new feature that allows their owners to summon them from parking spaces. Humans on foot pose another problem.

“We’ve learned that pedestrian behavior is somewhat erratic, so we want to keep a safe distance,” He said.

Chin said the company has had “zero incidents.” (A minor bump reported by the Brooklyn Eagle newspaper happened while one of Optimus’s vehicles was being driven manually, the company says.)

For all the questions yet to be resolved, Chin is keen to point out that today the company has a commercial service with paying customers. He declined to discuss the finances of his arrangement with Brookfield, saying only that it was a two-year agreement to provide three shuttles at the Reston site.

For Brookfield, a major real estate company with dozens of similar projects, there are two major attractions to working with Optimus: the potential its vehicles could provide developments with less space devoted to money-losing parking and a buzz of excitement from prospective tenants.

This state wants to usurp California as the capital of driverless cars

“They’re really intrigued,” said Greg Meyer, head of Brookfield’s Washington region. “I think most people recognize it’s the future.”

Academic experts and industry analysts say they expect low-speed autonomous vehicles to be more than novelties or stopgaps on the way to full-speed cars and will create new ways of getting around. They envision fleets connecting to bus and rail lines in big cities, where speeds on the road tend to be low anyway, and others designed to help the elderly and people with disabilities.

Coyner, the researcher, recently completed a study of low-speed vehicles for the Transportation Research Board, part of the National Academies of Sciences, Engineering, and Medicine. When she proposed the project in June 2018, Coyner said there were about a dozen projects underway, but by the time she wrapped up a few weeks ago, there were more like 225.

“Low speed is much more ready,” she said. “The interest in this exceeds what can be provided.”

In the Washington area, a company that makes a 3-D-printed shuttle has a facility at National Harbor and ran a pilot this summer at Joint Base Myer-Henderson Hall.

Early projects were simple tests, said Coyner, who works on mobility at the design firm Stantec, but routes are getting longer — one proposal in Hono­lulu would feature an 11-mile round trip — and operators are exploring more useful deployments that could complement existing transit.

A lot of cities want roboshuttles, including D.C. But will they work?

From Model T to driverless: Ford to launch fleet of robot cars in Washington

For major deployments on city streets, significant questions remain. Low-speed vehicles are not subject to the same safety standards as cars, potentially leaving them vulnerable in mixed traffic. And across the country, there is a patchwork of rules for autonomous vehicles, with federal regulators showing few signs of stepping in.

Chin recognizes the scale of the challenges and said the solution is to start with projects like the one in Reston.

“The real world here is pretty hard,” he said. “The real world out there is even harder, and the real world in Times Square is supremely difficult.”

“Our view is, we can’t do all of them,” he said. “Let’s start with places we can get autonomous vehicles to work reasonably in the near term.”